摘要
针对高校实验室传统安全管理中技防手段滞后、预警时效性差、数据孤岛等问题,该研究设计了一种基于物联网智能感知的实验室危险源监测系统,通过构建“硬件感知层-数据传输层-平台服务层-应用交互层”架构,集成多源传感器与人工智能(AI)视频监控设备,实现危险源实时动态监测。系统融合实验室分级分类管理、智能监测预警与数据可视化3个核心模块,建立“风险评价-阈值预警-多级响应-闭环处置”全流程管理体系,提升了实验室的安全性及管理效率,为高校构建智能化安全防控体系提供了技术范式。
[Objective]University laboratories are vital hubs for scientific research and talent development;however,they inherently contain diverse hazards,including flammable chemicals,high pressure,radiation,and complex equipment.Traditional safety management,which relies on manual inspections and isolated monitoring systems,often results in delayed warnings,coverage gaps,and inefficient response coordination,leading to persistent accident risks.Motivated by national policies promoting smart research infrastructure and the urgent need for proactive safety governance,an intelligent hazard monitoring system was designed and implemented in this study.This study transforms laboratory safety from a passive,reactive model to a dynamic prevention model by enabling real-time risk perception,intelligent early warning,and multilevel coordinated response,thereby safeguarding personnel and assets.[Methods]The system employs a meticulously designed four-layer internet of thing(IoT)architecture:①Hardware Perception Layer:Diverse sensors(electrochemical/nondispersive infrared/photoionization detector gas,temperature/humidity sensors,dust monitors,AI cameras with behavior recognition algorithms)are strategically deployed across laboratories,including concealed spaces such as fume hoods,to capture real-time environmental and visual data.②Data Transmission Layer:A dual-channel strategy ensures stable data flow.Sensor data is structured via serial servers and stored in MySQL,while AI camera streams are encoded(using H.264/AAC)and transmitted via RTMP/FLV to a streaming server(via HLS protocol),preventing interference in complex electromagnetic environments.③Platform Service Layer(Core Intelligence):This microservice-based layer integrates MySQL for structured data,Redis for caching,and the Camunda workflow engine.This layer is built on Spring Cloud Alibaba(Java).Key innovations include:Risk-based Dynamic Thresholds and Response:Laboratories are classified into Levels I-IV based on hazard severity and type,enabling the system to dynamically adjust alarm thresholds and responses.Intelligent Analysis:The system performs real-time fusion of multisource sensor data and video analytics to detect microleaks,abnormal behavior,and parameter coupling risks.Automated Multilevel Workflow:Camunda initiates a corresponding level of response when an alarm is triggered.This includes local sound/light alerts and escalating notification(via SMS,WeChat,or calls)from lab personnel to school administrators,while simultaneously retrieving relevant digital emergency procedures.④Application Interaction Layer:Vue3/ElementPlus(PC)and UniApp/uView(Mobile)interfaces provide a unified visualization of real-time data,video feeds,alarms,and device status,facilitating emergency process participation.Three core functional modules synergize on this architecture:①Laboratory Classification Engine:This module automates risk assessment(LevelsⅠ–Ⅳ)and categorization(e.g.,Chemical,Biological)based on hazards,providing the foundation for differentiated monitoring policies.②Intelligent Monitoring and Warning:This module leverages classification to deploy tailored sensor thresholds and AI video rules.It detects breaches and triggers the hierarchical alert and response workflow,and integrates a digital emergency procedure library.③Data Visualization Dashboard:Offers comprehensive"cockpit"views(real-time video,risk distribution maps,alarm stats,device health)for multidimensional analysis and decision support across room,building,and campus levels.[Results]Deployed across 249 high-risk laboratories(Levels I and II,including those handling explosive and toxic chemicals)at Xi'an Jiaotong University,the system demonstrated the following significant impact:①Enhanced Efficiency and Response:Automated classification and monitoring reduced safety inspection time by over 90%.The effectiveness of the system was validated by a critical real-world incident:at 3:27 AM,a CO microleak(58 ppm)was detected in a LevelⅡchemistry laboratory and instantly triggered Level 2 alerts(local alarms+multilevel notifications).Personnel confirmed the leak remotely via video feed,responded,and resolved the issue(caused by an unsecured valve)within 26 min,thereby preventing a potential explosion or poisoning.Post-analysis confirmed the leak would likely have gone undetected for hours without the system.②Robust 24/7 Protection:A network of 780 gas detectors,465 AI cameras,and 412 environmental sensors provided continuous coverage.User surveys(n≈300)indicated that 96.2%of staff and students felt significantly safer,particularly citing the reliability of the system during unmanned hours.③Optimized Coordination:The system automated over 2,400 laboratory information updates and facilitated seamless coordination across laboratory,department,and university levels.Of 1,482 valid alarms,21%required departmental intervention and 3.9%escalated to the university level,validating the effectiveness of the hierarchical workflow.④High Accuracy and Actionable Insights:The system demonstrated 96.61%accuracy(1,482 valid alarms out of 1,534 total).The primary triggers were humidity(67.27%,mainly summer highs prompting dehumidification),volatile organic compounds(21.12%,exposing procedural lapses such as spills or poor ventilation),and dust(4.59%,driving improved dust control).Alarm data is actively used to refine thresholds(e.g.,seasonal humidity adjustments,gas thresholds by lab level)and provides a quantitative foundation for environmental,health,and safety system development.[Conclusions]This IoT-based intelligent monitoring system represents a significant paradigm shift in the safety management of university laboratories.Its synergistic framework,which integrates multisource sensing,risk-based dynamic response,and data-driven visualization,enables real-time hazard perception,precise early warning,and efficient closed-loop incident management.Practical deployment confirmed substantial improvements in safety outcomes,operational efficiency,and collaborative emergency response.Although challenges such as sensor reliability in extreme environments and AI adaptability to new scenarios require ongoing mitigation,the system provides a scalable,proactive safety model through multisensor fusion,edge computing,and federated learning.Its"perception-decision-response"mechanism offers a valuable reference for future laboratory safety standards.Furthermore,integrating this system with access control,energy management,and campus security platforms promises the development of a holistic safety ecosystem underpinning high-quality scientific research.
作者
谷文媛
杨凡凡
李文武
王志飞
朱臻
GU Wenyuan;YANG Fanfan;LI Wenwu;WANG Zhifei;ZHU Zhen(Department of Laboratory Management,Xi'an Jiaotong University,Xi'an 710049,China;Office of Laboratory Management,Jilin University,Changchun 130012,China)
出处
《实验技术与管理》
北大核心
2025年第12期238-245,共8页
Experimental Technology and Management
基金
中国高等教育学会2023年度高等教育科学研究规划重大课题(23SYS0101)。
关键词
实验室安全
物联网
智能监测
分级预警
laboratory safety
internet of things
intelligent monitoring
graded early warning
