基于人工神经网络的大空间结构火场温度分析方法  

Temperature analysis method for large spatial structures under fire based on artificial neural networks

作　　者：王永宏 陈适才[1] 侯立群 WANG Yong-hong;CHEN Shi-cai;HOU Li-qun(Department of Architecture and Civil Engineering,Beijing University of Technology,Beijing 100124,China;China Railway Construction Group Co.Ltd.,Beijing 100049,China)

机构地区：[1]北京工业大学土木与建筑工程学院,北京100124 [2]中铁建设集团有限公司,北京100049

出　　处：《空间结构》2023年第3期31-39,共9页Spatial Structures

基　　金：国家自然科学基金项目(51678016).

摘　　要：为了简化结构火场温度边界条件,实现FDS火场模拟与后续的结构传热分析和热力耦合分析的无缝衔接,提出利用人工神经网络技术建立大空间火灾温度场预测分析模型的方法.根据大空间结构以及BP网络的特点,首先运用FDS分析大空间结构火灾场景;其次选择坐标和时间4个参数为输入向量,以火场温度为输出向量,建立BP神经网络模型;然后以FDS火灾模拟温度数据为样本,对网络模型进行训练和预测,建立大空间BP神经网络火灾温度场分析模型.结果表明,采用3层人工神经网络模型可以较好预测大空间结构的火灾温度场,预测结果温度场空间分布的相对误差在15%以内,温度场时程分析的相对误差在10%.In order to simplify the temperature boundary conditions of the structural fire field and realize the seamless connection between the FDS fire field simulation and the subsequent structural heat transfer analysis and thermal-mechanical coupling analysis,this paper proposes the application of artificial neural network technology to establish a large-space fire temperature field prediction and analysis model.Firstly,FDS fire field simulation software is adopted to simulate the large-space structure fire.Secondly,four parameters related to temperature field distribution are selected as input vectors,and the fire field temperature is taken as output vector to establish a BP neural network model.Lastly,taking FDS fire simulation temperature data as samples,the network model is trained,and the BP neural network fire temperature field prediction model is obtained.The results show that the three-layer artificial neural network model can predict the fire temperature field of large space structures.The relative error of the spatial distribution of the temperature field is less than 15%,and the relative error of the time history analysis of the temperature field is 10%.

关 键 词：大空间结构 空间温度场 火场模拟 BP神经网络 

分 类 号：TU393.3[建筑科学—结构工程]