小波函数的选择对结构损伤识别的影响  

Influence of the Selection of Wavelet Functions on Structural Damage Identification

作　　者：杨晓楠[1] 姜绍飞[2] 唐和生[1] 陈镕[1] 王远功[1] 

机构地区：[1]同济大学结构工程与防灾研究所,上海200092 [2]沈阳建筑大学土木工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2005年第6期635-639,共5页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金(50408033);辽宁省自然科学基金(20022136)

摘　　要：目的为了寻找出结构损伤识别中的最优小波函数,进而提高结构损伤的识别精度.方法选取了几种常用的小波函数,运用小波分析提取能量特征,采用小波概率神经网络方法,对四层钢框架进行了损伤识别及小波函数对损伤识别影响的对比研究.结果经过比较,发现采用函数曲线圆滑的Mexican cat小波函数识别精度可达100%,识别效果最好.结论在结构损伤识别领域,选用非正交、正则性好、消失矩阶数大的小波函数具有一定的优势.In order to obtain the optimum wavelet function and find the selecting criterion of wavelet functions on structural damage identification, wavelet transform（WT）was employed to extract energy feature of structural response. Several kinds of common wavelet functions were chosen to study the damage identifica- tion of a four-story steel frame by using wavelet probabilistic neural network method. The effect of different wavelet functions on damage identification was also discussed. This study shows that the smooth Mexican cat wavelet function is the best in structural damage identification（ Identification ratio = 100 % ）. Thus, the wavelet functions, which have the characteristics of non-orthogonality, good normalization and large orders of vanishing moment, have more predominance in the structural damage identification.

关 键 词：小波函数 能量特征 结构损伤识别 小波概率神经网络 钢框架 

分 类 号：TU318[建筑科学—结构工程] TU393