基于地形参数的地震动放大系数数值分析与预测方法  

Numerical analysis and prediction method of ground motion amplification factor based on topographic parameters

作　　者：罗麒锐 赵仕兴 吴启红 熊峰[4] 周巧玲 钟紫勤 杨姝姮 张敏[5] LUO Qirui;ZHAO Shixing;WU Qihong;XIONG Feng;ZHOU Qiaoling;ZHONG Ziqin;YANG Shuheng;ZHANG Min(Sichuan Provincial Architectural Design and Research Institute Co.,Ltd.,Chengdu 610017,China;Sichuan Engineering Research Center for Mechanical Properties and Engineering Technology of Unsaturated Soils,Chengdu 610106,China;Sichuan Provincial Engineering Research Center of City Solid Waste Energy and Building Materials Conversion and Utilization Technology,Chengdu 610106,China;College of Architecture and Environment,Sichuan University,Chengdu 610065,China;School of Civil Engineering and Geomatics,Southwest Petroleum University,Chengdu 610500,China)

机构地区：[1]四川省建筑设计研究院有限公司,成都610017 [2]非饱和土力学特性及工程技术四川省高校工程研究中心,成都610106 [3]四川省城市固废能源与建材转化技术工程研究中心,成都610106 [4]四川大学建筑与环境学院,成都610065 [5]西南石油大学土木工程与测绘学院,成都610500

出　　处：《建筑结构》2025年第2期50-57,35,共9页Building Structure

基　　金：四川省科技厅项目(2024JDRC0101);非饱和土力学特性及工程技术四川省高校工程研究中心开放基金项目(SC-FBHT2024-05、SC-FBHT2023-02);四川省城市固废能源与建材转化技术工程研究中心开放基金课题项目(GF2024ZD07);四川省建筑设计研究院院内科研项目(KYYN202313);西南石油大学自然科学“启航计划”(No.2023QHZ005)。

摘　　要：为复现真实地形,提出了基于图像识别技术的山脊地形建模方法,并进行了场地位置、高度、坡度和地形形状的参数分析。结果显示,相同高度和宽度的不同地形会表现出不同的地震响应,表明地形局部特征会影响地震动地形效应;在某些条件下,《建筑抗震设计规范》(GB 50011—2010)(2016年版)中最大地震动放大系数(GMAF)建议值1.6不足以使结构安全可靠。最后,基于大量仿真数据,建立了基于BP神经网络的GMAF预测模型。预测结果与仿真结果的对比显示,训练集与测试集的均方根误差仅为0.02和0.06,表明基于BP神经网络GMAF预测模型可用于结构抗震设计。In order to reproduce the real terrain,a ridge topography modeling method based on image recognition technology was proposed,and the parameters of the site location,height,slope and topographic shape were analyzed.The results show that different landforms with the same height and width exhibite different seismic responses,indicating that topographic local features would influence the ground motion topographic effect.In addition,under certain conditions,the maximum ground motion amplification factor(GMAF)recommended value of 1.6 in Code for seismic design of buildings(GB 50011—2010)(2016 edition)is not sufficient to make the structure safe and reliable.Finally,a GMAF prediction model based on BP neural network was established with a large number of simulation data.The comparison between the predicted results and the simulated results show that the root-mean-square error of the training set and the test set was only 0.02 and 0.06,which indicates that GMAF prediction model based on BP neural network could be used for structural seismic design.

关 键 词：地震动放大系数 地形效应 数值分析 图像识别 BP神经网络 

分 类 号：TU471[建筑科学—结构工程]