节点文献

WK-LS-SVM结构损伤模式识别方法的性能分析

Performance analysis of pattern identification based on WK - LS - SVM for structural damage

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 张茂雨李博王艳茹

【Author】 ZHANG Maoyu,LI Bo,WANG Yanru(Wenzhou University,Wenzhou 325035,China)

【机构】 温州大学

【摘要】 基于小波核函数最小二乘支持向量机方法(WK-LS-SVM)的研究成果,对WK-LS-SVM方法的抗噪声能力和适用性进行了分析。对10层框架结构的25种损伤工况进行了数值模拟,结果表明WK-LS-SVM方法具有良好的抗噪声能力。对不同结构相同损伤等级的自振频率进行分析,发现两者的自振频率变化率具有相似性,因此用某结构参数训练的支持向量机识别结构参数与之差异不大的结构的损伤情况,其识别误差小。因此实际工程中用简化估计的结构参数来识别损伤能满足工程精度要求。

【Abstract】 Based on the achievement in research on wavelet kernel funclion-least square-support vector machine(WK-LS-SVM) method,the anti-noise capability and adaptability of the method were analyzed.In this paper,25 typical damage conditions in a ten-storey frame structure were simulated.The results show that WK-LS SVM has a good anti-noise capacity.Further,the natural frequencies of different structures with same damage level were examined.The similarity of the change rate of natural frequencies was discovered.It shows that using supported vector machine trained with certain structural parameters to identify the damage level of similar structures is feasible and the induced error is small.Hence,identifying damage level of structure with simplified structure parameter estimation in practical engineering can satisfy engineering accuracy demand.

【基金】 温州市科技计划项目(S20100060)
  • 【文献出处】 世界地震工程 ,World Earthquake Engineering , 编辑部邮箱 ,2013年02期
  • 【分类号】TU312.3
  • 【下载频次】47
节点文献中: