摘要
提出了结合随机振动响应互相关函数、小波包分解和支持向量机(support vector machine,SVM)的结构损伤识别方法,计算了相邻测点响应的互相关函数幅值.采用小波包对得到的幅值进行分解,得到各个频带上的总能量;利用各频带上能量值存在的差异性作为输入到分类器的特征向量,训练SVM模型并对结构的损伤进行识别.应用该方法对Benchmark模型结构进行损伤判别,实验通过对比其他基于SVM的方法,结果表明该方法具有较好的识别精度.
A structural damage detection method by integrating cross-correlation function,wavelet packet decomposition （WPD） and support vector machine （SVM） was proposed. Cross-correlation functions amplitude were calculated on two acceleration responses which are obtained from two adjacent sensors. Then the processed signals were translate into energy features by WPD,the energy se quences at different bands of frequency were inputted to classifier as feature vectors. Finally,SVM as an effective classifier for small sample set problems was used to detect the multictass damage. The experiment results on a Benchmark model show that the proposed method obtained significantly higher identify accuracy than several other commonly used SVM-based methods.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第1期57-62,共6页
Journal of Xiamen University：Natural Science
基金
中央高校基本科研业务费(2010121065)
关键词
损伤识别
互相关函数
小波包分解
支持向量机
damage detectiom cross-correlation function~ wavelet packet decomposition~ support vector machine
