中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (1): 27-33.doi: 10.16265/j.cnki.issn1003-3033.2022.01.004

• 安全社会科学与安全管理 • 上一篇    下一篇

基于BP神经网络的建筑工人不安全行为预警模型

石娟1(), 常丁懿1, 郑鹏1,2,**()   

  1. 1天津理工大学 管理学院,天津 300384
    2华电电力科学研究院有限公司,浙江 杭州 310030
  • 收稿日期:2021-10-08 修回日期:2021-12-10 出版日期:2022-01-28 发布日期:2022-07-28
  • 通讯作者: 郑鹏
  • 作者简介:

    石娟(1978—),女,天津人,博士,教授,主要从事风险管理、安全工程等方面的研究。E-mail:
    石娟 教授,郑鹏 工程师

  • 基金资助:
    国家自然科学基金资助(71603181); 天津市教委社会科学重大项目(2021JWZD15); 天津市研究生科研创新项目(2021YJSB243)

An early warning model of unsafe behaviors of construction workers based on BP neural network

SHI Juan1(), CHANG Dingyi1, ZHENG Peng1,2,**()   

  1. 1School of Management, Tianjin University of Technology, Tianjin 300384, China
    2Huadian Electric Power Research Institute Co., Ltd., Hangzhou Zhejiang 310030, China
  • Received:2021-10-08 Revised:2021-12-10 Online:2022-01-28 Published:2022-07-28
  • Contact: ZHENG Peng

摘要:

为减少建筑工人不安全行为,提高企业安全管理水平,采用事故统计分析、文献分析、质性访谈方法获取不安全行为影响因素,从组织、个人、外在环境、设备4个方面建立不安全行为预警指标体系,在此基础上,基于反向传播(BP)神经网络原理,将预警指标作为网络输入,不安全行为无警、轻警、中警、重警4种状态作为网络输出,进而设计编制预警问卷,对问卷数据进行反复训练学习,最终构建出“23-9-4”3层结构的BP神经网络预警模型,并对该模型进行训练及测试。结果表明:该预警模型预警能力较强,能够较为准确地预测工人的不安全行为状态,从而可提前采取相应的防控措施。

关键词: 反向传播(BP)神经网络, 建筑工人, 不安全行为, 预警模型, 指标体系

Abstract:

In order to reduce unsafe behaviors of construction workers and improve safety management of corporations, methods of statistical analysis, literature analysis and qualitative interview were adopted to obtain influencing factors of unsafe behaviors. Then, an early warning index system was established from four aspects, namely organization, individual, environment and equipment. Based on BP neural network principle, with these indicators as network input, and four unsafe states were output, a warning questionnaire was designed, and the questionnaire data were repeatedly trained and learned. Finally, a three-layer BP neural network warning model of "23-9-4" was constructed, trained and tested. The results show that this model accurately predicts unsafe behavior states of workers, thereby enabling them to take prevention and control measures in advance.

Key words: back propagation (BP) neural network, construction workers, unsafe behavior, pre-warning model, index system