基于SEEW-SVM的结构损伤在线识别

Online identification of structural damage based on SEEW-SVM

在增量式特征向量加权支持向量机(WEVLS-SVM)结构损伤在线识别方法的基础上,提出了自适应特征向量指数加权向量机(SEEW-SVM)识别方法,该方法通过增加样本与修剪算法更新样本,并根据样本贡献量的大小对特征向量自适应进行指数加权。以剪切型结构为例进行了数值模拟分析,结果表明SEEW-SVM方法与WEVLS-SVM方法相比,不仅提高了识别精度,而且大大提高了识别效率,更适用于对结构的时变参数进行在线识别。

This paper presents the self-adaptive eigenvector exponential weighted Support Vector Machine（SEEW-SVM） based on the structural damage online identification of incremental weighted eigenvector Least Square Support Vector Machine（WEVLS-SVM）. This method updates the sample by incremental samples and pruned algorithm, and exponentialy weights the self adaption of eigenvector according to contribution of samples. The numerical simulations of shear building type demonstrates that in comparison with WEVLS-SVM method the proposed method greatly improves the identification precision and the computing efficiency, which adapts well to the change of structural parameters. Therefore, it is more suitable for online identification of structural damage.