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

作　　者：张茂雨[1] 李博[1] 王艳茹[1] 

机构地区：[1]温州大学,浙江温州325035

出　　处：《世界地震工程》2013年第2期151-155,共5页World Earthquake Engineering

基　　金：温州市科技计划项目(S20100060)

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。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.

关 键 词：结构损伤 模式识别 抗噪声能力 适用性 

分 类 号：TU317.9[建筑科学—结构工程] O327[理学—一般力学与力学基础]