结构可靠度计算的自适应辅助域方法

发布时间：2018-03-08 07:42

本文选题：结构可靠度　切入点：自适应　出处：《同济大学学报(自然科学版)》2017年04期 　论文类型：期刊论文

【摘要】：提出了一种Metropolis算法与支持向量机(SVM)方法相结合的自适应辅助域方法.利用Metropolis算法生成目标失效域内的条件样本点,并以该过程中的备选点作为初始训练点训练SVM模型.根据训练得到的SVM模型再自适应地选择一部分样本点加入训练点集,并对SVM模型进行更新,直至满足迭代终止条件.以最终得到的SVM模型作为辅助失效域,计算近似失效概率和两个条件失效概率.对近似失效概率进行修正,使最终得到的目标失效概率渐进无偏且更加稳定.算例表明该算法具有较好的计算精度、效率和鲁棒性.
[Abstract]:An adaptive auxiliary domain method combining Metropolis algorithm with support vector machine (SVM) method is proposed. The conditional sample points in the target failure domain are generated by using the Metropolis algorithm. According to the training SVM model, a part of the sample points are selected adaptively to join the training point set, and the SVM model is updated. Until the iterative termination condition is satisfied, the approximate failure probability and the two conditional failure probability are calculated using the final SVM model as the auxiliary failure region, and the approximate failure probability is modified. Finally, the final target failure probability is asymptotically unbiased and more stable. An example shows that the algorithm has good accuracy, efficiency and robustness.
【作者单位】：同济大学土木工程学院;
【基金】：国家自然科学基金(51261120374)
【分类号】：TB114.3

【相似文献】