基于支持向量机的高墩大跨桥梁体系可靠度计算

摘要

为解决高墩大跨桥梁结构体系可靠度求解过程中失效模式复杂、极限状态方程无法显式表达等问题,提出一种基于支持向量机(SVM)分类技术的体系可靠度计算方法.该方法采用拉丁超立方抽样法产生样本库,通过重复筛选构造一个精确的SVM分类器函数(而不是构造功能函数本身的响应面函数),然后采用蒙特卡罗法进行数值模拟计算体系失效概率.以济邵高速公路逢石河特大桥为例验证了采用该方法计算高墩大跨桥梁体系可靠度的实用性,计算结果表明该桥体系可靠度满足设计要求.%To cope with the problems that the failure modes in the solving process of structural system reliability of high-rise pier and long span bridge are complicated and the limit state e-quations can not be displayed and expressed, a method for computation of the system reliability based on the support vector machine (SVM) classification technique is proposed. The proposed method can generalize the sample bank, using the Latin hypercube sampling method and can construct a precise SVM classifier function (rather than the response surface function of the structure function itself) by repeated screening and the numerical simulation and computation of the system failure probability are accordingly carried out, using the Monte Carlo method. By way of example of the Fengshi River Bridge on Jiyuan-Shaoyuan Expressway, the practicability of the method for computation of the system reliability of the high-rise pier and long span bridge is verified and the results of the computation prove that the system reliability of the bridge can meet the design requirements