在城区中修建地铁隧道对安全性有极高的要求,传统的安全系数法虽然能在一定程度上反映工程的安全性,但无法考虑变量的随机性.针对传统子集模拟法的不足,引入了一种改进方法,对功能函数阀值处存在相等情况时的子集模拟法,提供一种解决思路.为解决子集模拟法在初始样本点少时的计算不准确问题,基于支持向量机的小样本学习能力,提出随机有限元法(SFEM)、支持向量机(SVM)和子集模拟法耦合的可靠度分析方法.使用耦合方法和传统方法分别计算衬砌单元的失效概率.并用经典算例与传统方法进行了对比,验证了其准确性和应用的可行性,可以作为样本点不足时分析地铁隧道的可靠度方法.结果表明,在样本数量较少的条件下,耦合算法具有很好的计算效率和精度,适用于隧道工程可靠度分析.%To construct subway tunnels in urban areas brings a heavy demand of safety. Although the traditional safety factor method can reflect engineering safety to some extent, it can not consider the randomness of variables. A modified subset simulation method is introduced to improve the insufficient of the traditional subset simulation method and a solution is presented to solve the problem when there are a number of values equaling to threshold value of performance function. A reliability analysis method combining the subset simulation method with stochastic finite element method and support vector machine is presented, which takes full advantage of support vector machine, overcoming the inaccurate of traditional subset simulation method when the initial sample points is not enough. This paper calculates the failure probability of tunnel lining by the new method and the traditional methods respectively and compares the results. Comparing with the classical sample, this paper verifies the accuracy and the feasibility of the new method, so as to be used as a reliability analysis method for subway tunnel. In this example, the proposed method is more fitting for engineering contexts for a small size of samples.
展开▼
