基于BP神经网络的RC框架结构主余震易损性分析  

Vulnerability Analysis of the RC Framework Structure Subjected to the Mainshock-aftershock Sequence Based on BP Neural Network

作　　者：邓夕胜[1] 周紫娟 赖馨粤 林嘉聪 朱一林 DENG Xisheng;ZHOU Zijuan;LAI Xinyue;LIN Jiacong;ZHU Yilin(School of Civil Engineering and Geomatics,Southwest Petroleum University,Chengdu 610500,Sichuan,China;School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,Sichuan,China;Fujian No.1 Construction Group Co.,Ltd.,Sanming 365001,Fujian,China)

机构地区：[1]西南石油大学土木工程与测绘学院,四川成都610500 [2]西南交通大学土木工程学院,四川成都610031 [3]福建一建集团有限公司,福建三明365001

出　　处：《地震研究》2025年第3期496-506,共11页Journal of Seismological Research

基　　金：国家自然科学基金面上项目(12372143);广元城投集团科技基金(GYCT-KY-202101).

摘　　要：为研究RC框架结构在主余震序列地震动作用下的易损性,设计了一个6层RC框架结构,挑选不同余震持时的主余震地震动合成得到主震-长持时余震和主震-短持时余震序列地震动各20组,选用地震动强度指标作为预测变量,将全部地震序列输入到结构中进行IDA分析,以IDA分析结果作为神经网络数据库,训练出最佳BP神经网络模型,得到BP神经网络的结构易损性曲面,从而综合考虑余震持时和主震强度的向量型IM与结构最大层间位移角之间的关系。结果表明:在主震-长持时余震和主震-短持时余震两种地震作用下,前者对结构的影响较大,造成结构的失效概率更大;基于BP神经网络的易损性函数能更好地反映结构损伤。In order to study the vulnerability of the RC frame structure under the action of the mainshock-aftershock earthquake sequence,we design a 6-layer RC frame structure.By selecting recordings of the ground motion of the mainshock-aftershock sequence containing aftershocks with different durations,we respectively synthesize 20 sets of mainshock-aftershock sequence containing long-duration aftershocks and 20 sets of mainshock-aftershock sequence containing short-duration aftershocks.We select the intensity index of the ground motion IM as the predictive variable,and input all the ground motions of the 40 sets of mainshock-aftershock sequence into the structure for IDA analysis.Then we take the IDA analysis results as the neural network database to train the best BP neural network model.Thus,we get the vulnerable surface of the BP neural network.On this basis we obtain the relationship between the IM and the maximum interlayer displacement angle of the structure.We find that the mainshock-aftershock sequences containing long-duration aftershocks impacts the structure greater than the ones containing short-duration aftershocks.The vulnerability function based on BP neural network can better reveal the damage to the structure.

关 键 词：主余震 持时 BP神经网络 RC框架结构 易损性曲面 

分 类 号：P315.92[天文地球—地震学]