基于小波包变换的支持向量机损伤诊断方法  

Diagnosis of Structure Damage by Support Vector Machine Based on Wavelet Packet Transform

作　　者：赵学风[1] 段晨东[2] 刘义艳[2] 韩旻[2] 

机构地区：[1]西安交通大学电气工程学院,西安710049 [2]长安大学建工学院,西安710061

出　　处：《振动．测试与诊断》2008年第2期104-107,共4页Journal of Vibration,Measurement & Diagnosis

基　　金：陕西省自然科学基金资助项目(编号:2005E205)

摘　　要：针对结构损伤识别中缺少实际损伤样本的问题,提出基于小波包特征提取的支持向量机结构损伤诊断方法。该方法将结构振动信号小波包分解后的频带能量,经过多传感器数据融合后作为特征向量,输入到多分类的支持向量机中,实现了结构多损伤的识别和定位。应用该方法对IASC-ASCE模型进行了分析,试验结果表明,小波包分解频带能量能够较好地反映结构的损伤特征。多传感器数据融合能够使不同传感器的信息相互补充,减小了损伤检测信息的不确定性,提高了损伤诊断准确率。For lack of actual damage samples in the structure damage diagnosis,a method called support vector machines(SVM) based on the feature extraction by wavelet packet decomposition is proposed.The energy sequences at different bands of frequency decomposed by the wavelet packet transformation are fused to form extracted feature vectors,which are inputted to a multi-classified SVM to implement multi-damage recognition and damage localization.Tested and analyzed with the IASC-ASCE model by using the method,it proves that the wavelet packet energy sequence could reflect the damage condition,the data fusion could enrich the diagnosis information and reduce the uncertainty of the damage detection information,and the diagnosis accuracy is improved.

关 键 词：结构损伤诊断 能量序列 支持向量机 数据融合 特征提取 

分 类 号：O327[理学—一般力学与力学基础]