济南黄河隧道智慧防灾系统架构研究及验证
Research and validation of the intelligent disaster prevention system for the Jinan Yellow River Tunnel
针对长大穿黄隧道环境状态难以综合感知、火灾发展态势难以动态掌握以及应急疏散决策依赖固定预案等问题,提出一种面向复杂隧道场景的智慧防灾方法,构建由环境参数感知、火情监控预警、火灾态势重构和动态应急疏散组成的隧道智慧防灾系统。首先,融合隧道环境监测数据与短时交通预测结果,建立基于模糊控制的综合污染评价与通风辅助调控方法。其次,构建基于多源监测数据融合的火情监控预警模型,利用 GA-BP(genetic algorithm-backpropagation)神经网络识别明火、阴燃火及非火灾状态,并结合分级决策方法形成火情预警指令。在此基础上,利用长短期记忆网络挖掘温度时序数据与火灾演化参数之间的映射关系,动态预测火源位置和热释放速率等关键火情信息。进一步结合多智能体疏散仿真方法,综合考虑烟气、能见度、有毒气体及人员行为等因素,生成随火场环境和人员分布动态更新的疏散方案。以济南黄河隧道为工程背景开展系统构建与验证。结果表明,火情监控预警模型在典型模拟工况下较传统温差阈值方法提前约 20 s 达到报警判据,LSTM(long short-term memory)模型在温度数据完整和部分高温测点数据缺失条件下,热释放速率(heat release rate,HRR)预测决定系数分别达到 95.9%和 93.7%,16 组典型疏散工况的总疏散时间为 654~780 s。研究结果表明,该系统能够实现环境状态感知、火情预警、态势重构与动态疏散决策的连续联动,可为复杂隧道智慧防灾系统建设和火灾应急管理提供工程参考。
An intelligent disaster prevention system for complex tunnel scenarios was developed to improve environmental perception, fire monitoring, situation reconstruction, and emergency evacuation in long underwater river-crossing tunnels. The system consisted of four functional modules: environmental-parameter perception, fire monitoring and early warning, fire situation reconstruction, and dynamic evacuation. Environmental monitoring and ventilation regulation were conducted using a fuzzy-control-based method. A GA-BP (genetic algorithm-backpropagation) neural network was employed for fire early warning through multi-source data fusion, while a long short-term memory (LSTM) network was used to predict the fire source location and heat release rate (HRR). A multi-agent model was introduced to generate dynamic evacuation strategies, with smoke propagation, visibility, toxic gases, and occupant movement characteristics taken into consideration. The proposed system was validated using the Jinan Yellow River Tunnel as the engineering background. The results showed that the fire warning model reached the warning criterion approximately 20 s earlier than the conventional temperature-difference threshold method. The LSTM model achieved coefficients of determination (R2) of 95.9% and 93.7% for HRR prediction under complete data and partial high-temperature sensor failure conditions, respectively. The total evacuation time for 16 typical evacuation scenarios ranged from 654 to 780 s. The effectiveness of the proposed system in continuous and coordinated environmental perception, fire warning, situation reconstruction, and dynamic evacuation decision-making was demonstrated through the validation. The proposed framework was considered to provide an engineering reference for intelligent disaster prevention and emergency management in complex tunnels.
| [1] |
郭志国, 李益锌, 周令剑, |
| [2] |
|
| [3] |
李岩, 蓝翎源, 单宇, |
| [4] |
|
| [5] |
于海成, 田羽, 李庆健, |
| [6] |
|
| [7] |
杨淞成, 鲍玉朋, 赵兰明. 基于黏度特征气体探测识别方法实验研究[J]. 消防科学与技术, 2020, 39(2): 243-246. |
| [8] |
|
| [9] |
胡燕, 王慧琴, 张国飞, |
| [10] |
|
| [11] |
颜礼彬. 基于独立分量分析和颜色特征的视频火焰图像分割法[J]. 传感器与微系统, 2022, 41(4): 149-152. |
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
雷文君, 郭丽丽, 赵旭明, |
| [17] |
|
| [18] |
|
| [19] |
|
| [20] |
黄超. 平安隧道救援站排烟模式与人员疏散研究[D]. 成都: 西南交通大学, 2015: 51-60. |
| [21] |
|
| [22] |
王明年, 李琦, 于丽. 长大铁路隧道紧急救援站疏散设施设计参数研究[J]. 隧道建设, 2017, 37(3): 285-290. |
| [23] |
|
| [24] |
樊雯婧. 井下火灾多救护队救援路径优化的群智能算法研究[D]. 西安: 西安建筑科技大学, 2014: 11-40. |
| [25] |
|
| [26] |
黄莺, 郑雪琳, 赵鑫. 基于改进蚁群算法的展览馆火灾疏散路径研究[J]. 消防科学与技术, 2023, 42(10): 1362-1369. |
| [27] |
|
| [28] |
JTGD70/2-2014 公路隧道设计规范第二册交通工程与附属设施[S]. |
| [29] |
JTG/T 3661-2025 公路隧道交通工程与附属设施施工技术规范[S]. |
| [30] |
何应道, 郭超, 沈奕, |
| [31] |
|
济南济泺路穿黄隧道工程科研资助项目(TSYSDY-JLLCHSD[2020]-07)
/
| 〈 |
|
〉 |