基于BP神经网络的单层钢筋混凝土柱工业厂房震害预测  

Seismic damage prediction for single-story reinforced concrete industrial building based on back propagation neural network

作　　者：赵艳林[1] 杨军平[1] 黄剑飞 吕海波[3] 

机构地区：[1]桂林工学院土木工程系,广西桂林541004 [2]广西南宁市建筑设计院,南宁5300122 [3]广西大学土木建筑工程学院防灾减灾研究所,南宁530004

出　　处：《桂林工学院学报》2006年第4期491-496,共6页Journal of Guilin University of Technology

基　　金：广西科技攻关项目(桂科攻0480003);广西教育厅重点科研项目(桂教科研(2003)22)

摘　　要：将人工神经网络理论应用于等高单层钢筋混凝土柱工业厂房的震害预测.在分析震害特点的基础上,将震害影响因子分为精确性和规律性两大类,提出以地震反应指标、天窗类型、支撑情况、建筑材料作为主要的影响因子,并给出了相应的量化取值范围,然后将震害等级作为输出结果,构造了震害预测的BP人工神经网络.通过对52个实际震害实例的检验,网络的准确率超过80%.计算结果证明了该人工神经网络的有效性.A back propagation artificial neural network is applied to predict seismic damage of single-story reinforced concrete industrial building. Based on the analysis of characteristics of seismic damage, it is found that earthquake response index, type of skylight, bracing system and building material are the main factors affecting seismic damage. The four factors can be classified into two types： precise factors and regular factors. The corresponding spans of factors are suggested and applied to engineering examples. Thus the back propagation artificial neural network is developed, with factors affecting seismic damage as input and seismic damage grade as output. Verified by 52 engineering examples, the percentage of accuracy is above 80%. It is concluded that the back propagation artificial neural network developed in this paper is applicable to predict seismic damage of single-story reinforced concrete industrial building.

关 键 词：工业厂房 震害预测 人工神经网络 

分 类 号：TU352.1[建筑科学—结构工程]