中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (4): 155-162.doi: 10.16265/j.cnki.issn1003-3033.2022.04.023

• 公共安全 • 上一篇    下一篇

大数据视域下体育场馆动态火灾风险指标研究

卢颖1,2(), 赵志攀1, 姜学鹏1,2, 吴锦东3, 范小鹏1   

  1. 1 武汉科技大学 资源与环境工程学院,湖北 武汉 430081
    2 湖北省工业安全工程技术研究中心,湖北 武汉 430081
    3 武汉理工光科股份有限公司,湖北 武汉 430070
  • 收稿日期:2022-01-18 修回日期:2022-03-25 出版日期:2022-04-28 发布日期:2022-10-28
  • 作者简介:

    卢颖 (1986—),女,湖北钟祥人,博士,讲师,主要从事城市公共安全风险理论与控制技术、消防安全管理等方面的研究。E-mail:

    姜学鹏, 教授。

  • 基金资助:
    国家自然科学基金资助(51874213); 湖北省自然科学基金资助(2018CFB186); 湖北省应急管理厅安全生产专项项目(KJZX201907011)

Dynamic fire risk indexes for stadiums from perspective of big data

LU Ying1,2(), ZHAO Zhipan1, JIANG Xuepeng1,2, WU Jindong3, FAN Xiaopeng1   

  1. 1 School of Resource and Environmental Engineering, Wuhan University of Science and Technology, Wuhan Hubei 430081, China
    2 Hubei Provincial Industrial Safety Engineering Technology Research Center, Wuhan Hubei 430081, China
    3 Wuhan University of Technology and Optical Science Co., Ltd., Wuhan Hubei 430070, China
  • Received:2022-01-18 Revised:2022-03-25 Online:2022-04-28 Published:2022-10-28

摘要:

为解决体育场馆火灾风险评估多采用静态指标而导致动态指标不明确,且动态评估所需物联网监测数据具有多样性和复杂性的问题,分析消防主机、消防水箱液位等48种物联网监测数据特征,构建消防主机故障点位占比、实际与标准液位差等可量化的动态指标体系;建立27个体育场馆监测数据的指标数据集,运用随机森林算法筛选和优化48个指标,研究动态火灾风险评估指标的构建与优化。结果表明:在删除重要度最低的9维特征时,均方误差达到最低0.05,获得最优的体育场馆动态火灾风险评估指标体系。

关键词: 大数据, 体育场馆, 动态火灾风险, 火灾风险评估, 随机森林算法

Abstract:

In order to solve problems that static indicators are more frequently used in fire risk assessment of stadiums, while dynamic indicators are not clear, and internet of things monitoring data required for dynamic assessment is diverse and complex, characteristics of 48 kinds of internet of things monitoring data such as fire host and fire tank liquid level were analyzed, and a quantifiable dynamic index system was constructed, including fault location percentage of fire hosts and difference between actual and standard liquid level. Then, an data set based on monitoring data of 27 stadiums was established, 48 indicators were screened and optimized using random forest algorithm, and development and optimization of dynamic fire risk assessment indicators were studied. The results show that when the 9-dimensional features with the lowest importance are deleted, mean square error reaches the minimum of 0.05, and optimal dynamic fire risk assessment index system for stadiums is obtained.

Key words: big data, stadium, dynamic fire risk, fire risk assessment, random forest algorithm