基于能量特征的小波概率神经网络损伤识别方法  

Damage identification methods of wavelet probabilistic neural network based on energy features

作　　者：杨晓楠[1] 姜绍飞[2] 王金鱼[2] 

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：《兰州理工大学学报》2005年第3期123-126,共4页Journal of Lanzhou University of Technology

基　　金：国家"十五"科技攻关(2002BA806B4);国家自然科学基金(50408033);建设部科技项目(0221.3);辽宁省自然科学基金(20022136)

摘　　要：以小波能量特征向量作为概率神经网络(PNN)的输入向量集,提出了小波概率神经网络(WPNN)的损伤识别方法.为了验证该方法的有效性,对钢框架进行了损伤识别研究,并考虑了随机噪声的影响.识别结果表明:WPNN抗噪声能力强,识别精度高,在结构损伤识别与在线检测方面具有潜力.By combining wavelet energy feature vectors with probabilistic neural network (PNN) in noisy conditions,a new damage identification method called wavelet probabilistic neural network (WPNN) was proposed.Damage identification of a steel frame was utilized to illustrate the effect of this method,and the noise was also considered.The identification result showed that it has high identification accuracy and noise-resistant,being of potential in on-line structural damage detection.

关 键 词：多小波变换 能量特征 结构损伤识别 小波概率神经网络 框架结构 

分 类 号：TU318[建筑科学—结构工程] TU393