结构体系可靠度分析的改进支持向量回归  

Adaptive support vector regression method for structural system reliability analysis

作　　者：刘扬[1] 鲁乃唯[1,2] 蒋友宝[1] 

机构地区：[1]长沙理工大学土木与建筑学院,湖南长沙410114 [2]东南大学土木工程学院,江苏南京210096

出　　处：《浙江大学学报（工学版）》2015年第9期1692-1699,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家"973"重点基础研究发展规划资助项目(2015CB057700);国家自然科学基金资助项目(51378081;51108045;11102029)

摘　　要：针对结构体系可靠度分析的关键问题,在传统支持向量回归方法(SVR)的基础上引入2个支持向量更新步骤,提出适用于结构体系可靠度分析的改进支持向量回归(ASVR)方法.将传统SVR与遗传算法(GA)结合搜索关键构件功能函数的验算点,采用重要抽样Monte-Carlo模拟得出构件的失效概率.基于β约界法筛选潜在的失效构件,根据变化的有限元模型再次更新SVR模型.基于上述2个更新步骤,改进传统构件可靠度分析的SVR算法并用于结构体系可靠度分析.数值算例分析表明该算法的计算效率与精确性,斜拉桥的算例分析证实该算法在实际工程结构中的适用性,同时得到该斜拉桥的2条主要失效路径.Two updating processes of support vectors were introduced into the traditional support vector regression（SVR）method for the purpose of presenting a new adaptive support vector regression（ASVR）method,in order to solve the key issues of structural system reliability assessment.Importance sampling Monte-Carlo simulation（MCS）was used to estimate structural reliability after searching the checking point of key components functions based on traditional SVR and genetic algorithm（GA）.Screen the potential failure components using theβ-bound method,and update the SVR models according to the updated finite element models.With the two updating processes,the traditional SVR was improved and applied to structural system reliability analysis.Numerical examples were given to illustrate the accuracy and efficiency of the proposed method.The numerical analysis of cable-stayed bridge demonstrates the applicability of ASVR method in the practical engineering structures.Meanwhile,two main failure sequences of the cable-stayed bridge were deduced.

关 键 词：结构体系可靠度 支持向量回归 极限状态方程 失效路径 MONTE-CARLO方法 遗传算法(GA) 

分 类 号：TU311.4[建筑科学—结构工程]