基于建筑部件行为推理与环境感知约束的室内火灾应急场景动态构建与知识表达研究

This project aims at enhancing the cognition of complex changes in building structures and environmental information in an indoor fire emergency. Starting from the perspective of safe evacuation, this project first proposes a three-dimensional (3D) semantic building model by taking into account the dynamic behavior reasoning of building components. The model extends the thematic semantic information of an indoor fire emergency and considers "fire districts - building components - building entrances "as a basis for the 3D topological representation of semantic association. The proposed model can improve the capability for the dynamic analysis of interior spaces and structural components in a fired building. Using the semantic components as associated carriers, this project further develops a multi-level context-ware method associated with the integration of multi-source sensors and indoor fire simulations. It facilitates the analysis and predictions of evacuations under the conditions occurring in different stages of fire. Based on behavior reasoning of building components and the perception of dynamic environments, this project finally develops dynamic knowledge representations of multivariate spatio-temporal associations. The knowledge representation breaks through the traditional information communication using means of "node-path-network ", leading to an easy understanding for policy makers. The result will reveal the dynamic characteristics of interior spaces and flue gas environment in a fired building. It has a potential to provide effective support for spatial analysis, real-time guidance and dynamic decision-making for evacuation in a fire emergency.

针对建筑火灾发生时的高动态认知难题，本项目从安全疏散空间动态判别和建筑烟气环境的动态感知角度出发，首先拓展建筑火灾应急的专题语义信息，基于"防火分区-建筑部件-建筑开口"为语义关联基准的三维空间拓扑表达，建立顾及建筑部件动态行为推理的三维建筑语义信息模型，发展面向设施有效性与空间通达性的建筑部件动态行为推理方法，实现建筑物内部空间与部件的可动态分析能力；进而以语义部件为信息关联载体，发展融合多源传感器与室内火灾模拟的多层次环境感知方法，力求支持建筑火灾不同发展阶段影响下的实时疏散分析与预测；最后综合决策者与逃生者的双向视角，突破"节点-路径-联通网络"的信息传递模式，发展基于部件行为推理与动态环境感知约束的多变量时空关联模式与动态知识表示方法。研究结果对综合揭示建筑空间与烟气环境的动态特征有所推进，并可望为火灾发生时的逃生空间分析以及实时疏散引导与动态决策提供有效支撑。

建筑火灾发生时有其突发性、多变性和烟毒性等特点，复杂的室内空间结构更可能使人们不能冷静的采取应对方式，丧失扑救与逃生的第一时间。室内火灾的高动态特征至使逃生人员与决策者所关注的焦点正逐步从整体、静态、抽象转到室内局部、动态、细致。如何充分认知火灾发生时的建筑结构空间和建筑烟气环境，从而动态识别并利用有限的建筑逃生空间，提供快速准确的应急疏散和救援已经成为亟待解决的关键科学问题之一。.本研究立足于安全疏散空间动态判别和建筑烟气环境的动态感知，首先深度剖析与扩展面向火灾应急的关键特征（如走廊、楼梯、电梯、防火门、楼梯间、防火分区等）及多层次语义关系，并重点分析室内火灾的高动态环境变化引发的建筑部件（如防火门/防火卷帘）功能、状态与生命周期的改变，建立顾及建筑部件动态行为推理的三维建筑语义信息模型，力求推动建筑物内部空间与部件的可动态分析能力，为实现疏散路径安全与可通达空间的定量化提供模型基础。其次，结合固定传感节点或移动传感设备的实时动态监测室内空间的微环境变化，并主动集成火灾过程（FDS）模拟预测结果，在揭示高层建筑内火灾烟气输运和特征参数的时空分布规律的基础上，发展多层次环境感知方法，力求实现火灾过程中的某一阶段或时刻的内部结构与微环境状态（突发，发展、轰然和蔓延至整个防火分区）的动态感知。最后，综合考虑决策者与逃生者的双向视角，发展面向火灾模拟过程的多变量关联模式与综合表达，为实时疏散引导与动态决策提供支撑。项目执行过程中，发表学术论文11篇，其中SCI/SSCI论文6篇，EI论文1篇，参加国内外重要学术会议8人次，结合项目主要研究内容培养博士研究生2名。