地铁施工动态安全知识地图及可视化研究

董超

摘 要

近年来，我国地铁施工安全事故数仍然居高不下，造成了巨大的生命和财产损失。安全知识的利用不足严重制约了地铁施工安全管理水平的提升，如何识别在地铁施工中安全知识利用不足的原因，进而利用在信息化背景下产生的各类安全数据来解决安全知识重用的问题，已成为地铁施工安全管理中的一个亟待解决的重要课题。

本文借鉴国内外相关研究，针对武汉地铁施工工程，提出了地铁施工动态安全知识地图及可视化研究，实现了对地铁施工安全知识流的可视化，识别出了安全知识流中存在的障碍；并综合运用非参数贝叶斯网络、位置概率网格、Copula 模型等方法，实现了基于多种数据源的安全知识可视化，消除了知识流障碍对安全知识重用的制约。

1、本文以武汉地铁施工安全风险管理为背景，构建了地铁施工动态安全知识地图，实现了对地铁施工安全知识流的可视化，识别出安全知识特征障碍、知识接收者障碍、语境障碍和机制障碍是造成安全知识不能充分重用的主要因素。

2、针对安全知识特征障碍，本文构建了基于专家经验的地铁盾构施工安全风险评估模型。以结构化专家访谈提取出风险因素自身不确定性分布，及风险因素间的积距相关系数；并以武汉地铁四号线复兴路站险情的调查数据，验证了模型的有效性；最后以安全风险图完成了基于专家经验的安全知识可视化。

3、针对知识接收者障碍和语境障碍，本文构建了基于位置概率网格的工人前瞻性风险评估模型。知识接收者障碍和语境障碍影响了工人对安全知识的吸收，以致不能够对自身的安全形势进行及时有效的评估。本文以位置概率网格对工人的运动进行建模，实现了基于工人下一个位置预测的前瞻性风险评估及智能预警，并以安全风险灰度图完成了基于空间数据的安全知识可视化。

4、针对安全知识流中的机制障碍，构建了基于未遂事件时序数据的D-VineCopula 模型，揭示了蕴藏在安全数据背后的时间自相关性，消除了安全数据非公开性对安全知识流的阻碍，并以地铁施工安全风险随未遂事件月发生数量的变化图完成了基于时间数据的安全知识可视化。保障安全知识的充分利用是提升地铁施工安全管理水平，减少安全事故发生的重要环节。本文构建了地铁施工动态安全知识地图，地铁盾构施工安全风险评估模型，地铁施工工人前瞻性安全风险评估模型，地铁施工未遂事件时序数据D-VineCopula 模型，实现了对地铁施工安全知识流及基于不同类型数据源的安全知识可视化，促进了安全知识在地铁施工安全管理组织中的流动，保障了安全知识的充分利用。 通过实际工程案例，验证了模型的有效性，对于提升地体施工安全管理水平具有重要意义。

关键词： 地铁施工安全，知识可视化，动态知识地图，贝叶斯网络，位置概率网格，Copula模型

Abstract

Statistical data on accident rates associated with metro construction projects indicated the high risks involved, which caused a huge loss of life and property. The circumstance, that project safety knowledge cannot be made full reuse, has restricted the improvement of metro construction safety management. How to identify the reasons why the safety knowledge cannot be made full reuse in metro construction, and how to solve the problems based on generated safety data, has become important issues in metro construction safety management.

This dissertation put forward the study on safety knowledge-dynamics integrated map and visualization for metro construction, which facilitated the visualization of safety knowledge flow in the metro construction safety management; and the Non-parametric Bayesian network, position probability grid, and Copula have been adopted in this dissertation, to visualize the safety knowledge based on different types of data sources, which facilitated eliminating safety knowledge flow barriers in metro construction.

First, this dissertation took the Wuhan Metro Project as the research object and constructed the safety knowledge-dynamics integrated map for metro construction, which visualized the safety knowledge flow in metro construction. Then, the knowledge flow barriers have been identified, which were knowledge characteristics barrier, the knowledge receiver barrier, the context barrier and the mechanism barrier.

Second, this dissertation constructed the safety risk assessment model for metro shield construction based expert experience, which was to eliminate the knowledge characteristics barrier. Structutered expert judement was adopted to extract the uncertainty distribution of risk factors and the product correlation coefficient between risk factors. The validity of the model has been verified by the survey data of Fuxing Road Station incident of Wuhan Metro Line 4. Finally, the safety knowledge visualization based on epert experience has been completed by the safety risk map.

Third, this dissertation constructed the workers’ proactive risk detection model based on the position probability grid, which was to eliminate the knowledge receiver barrier and context barrier. The knowledge receiver barrier and context barrier affected the workers’ absorption of safety knowledge, which made it impossible to identify their safety situations timely and effectively. This dissertation modeled the workers’ movement by the position probability grid, and realized the proactive risk assessment and intelligent early warning based on the workers’ next location prediction. The safety knowledge visualization based on spatial data was completed by the gray-scale risk map.

Finally, this dissertation constructed the D-vine Copula model based on the near-miss time series, which revealed the time autocorrelation behind the safety data. It eliminated the safety knowledge flow mechanism barrier caused by the non-disclosure of safety data, and completed the safety knowledge visualization based on time data with the safety risk map along the number of near-misses each month.

Ensuring the full reuse of safety knowledge is of great significance for improving the metro construction safety management and reduce the occurrence of safety accidents. This dissertation constructed the safety knowledge-dynamics integrated map of metro construction, the safety risk assesstment model for metro shield construction, the proactive risk detection model for metro construction workers and the D-vine Copula model of near-miss time series. The visualization of safety knowledge flow and safety knowledge based on different types of data sources has been realized, which promoted the reuse of safety knowledge in metro construction. The validity of the models has been verified by practical engineering cases, which is of great significance for improving the metro construction safety management.

Key words：metro construction safety, knowledge visualization, knowledge-dynamics integrated map, Bayesian networks, Copula