基于高斯过程回归的桥梁多变量地震易损性分析  

Multivariate seismic vulnerability analysis of bridges based on Gaussian process regression

作　　者：闫业祥 孙利民[1,2,3] YAN Yexiang;SUN Limin(Department of Bridge Engineering,Tongji University,Shanghai 200092,China;State Key Lab for Civil Engineering Disaster Prevention,Tongji University,Shanghai 200092,China;Shanghai Qizhi Research Institute,Shanghai 200232,China)

机构地区：[1]同济大学桥梁工程系,上海200092 [2]同济大学土木工程防灾国家重点实验室,上海200092 [3]上海期智研究院,上海200232

出　　处：《振动与冲击》2022年第23期27-35,共9页Journal of Vibration and Shock

基　　金：国家重点研发计划(2017YFC1500605)。

摘　　要：建立了一种基于高斯过程回归的多变量地震易损性分析方法,可以将易损性表达为由地震动强度指标和结构参数组成的矢量输入的多元函数,并以此为基础发展了均值易损性曲线与包络易损性曲线,能够定量分析所有不确定参数对易损性的影响。以一座三跨钢筋混凝土连续箱梁桥为例进行了分析与验证,结果表明:所提出的方法能够以数百次的动力分析达到与蒙特卡罗法(数十万次动力分析)相一致的精度;并识别了不同的地震强度指标对易损性的影响,发现任何单一的参数均会造成易损性的不精确(离散)呈现;对于该算例,结果显示平均谱加速度是一个用来生成最小离散度的均值与包络易损性曲线的最优指标。Here, a multi-variable seismic vulnerability analysis method based on Gaussian process regression was proposed. With this method, vulnerability could be expressed as a multivariate function of vector input composed of ground motion intensity index and structural parameters. Based on this function, the mean vulnerability curve and envelope vulnerability curve were developed, they could be used to quantitatively analyze effects of all uncertain parameters on vulnerability. Taking a 3-span reinforced concrete continuous box girder bridge as an example, analysis and verification were performed using the proposed method. the results showed that the proposed method through hundreds of dynamic analyses can reach the accuracy consistent to that of Monte Carlo method through hundreds of thousands of dynamic analyses;any single parameter can cause imprecise or discrete presentation of vulnerability;for the example above mentioned, the average spectral acceleration is the optimal index to generate mean vulnerability curve and envelope vulnerability one with the minimum dispersion.

关 键 词：多变量地震易损性分析 均值易损性曲线 包络易损性曲线 高斯过程回归 钢筋混凝土连续梁桥 

分 类 号：U442.55[建筑科学—桥梁与隧道工程] U447[交通运输工程—道路与铁道工程]