摘要
基于小波包分析和神经网络的声发射信号缺陷检测方法,提出采用区间小波包分解与能量距相结合作为声发射信号的特征向量,取代了传统的"小波包-能量"特征提取方法,并以金属罐形容器罐底缺陷诊断为例验证了该方法的有效性.结果表明,基于区间小波包能量距的神经网络特征提取方法更好地利用了缺陷信号的主要频带和小波包分析的时频信息,与传统方法相比,能大大简化检测系统的复杂度,提高容器的检测识别率.
In the fault diagnosis through acoustic emission technique based on wavelet analysis and neural networks,the wavelet packet method based on sections and energy-moment feature was used to replace the traditional 'wavelet packet-energy' to pick-up characteristics of AE signals.Then the efficiency of this method was validated in the fault diagnosis of the bottom of metal vessel.The result shows that,compared with ordinary method,the pick-up method based on wavelet packet of sections and energy-moment feature c...
出处
《山东理工大学学报（自然科学版）》
CAS
2008年第6期96-98,104,共4页
Journal of Shandong University of Technology:Natural Science Edition
基金
山东理工大学科技计划资助项目(2006KJM14)
关键词
声发射
区间小波包
能量距
神经网络
特征提取
acoustic emission
wavelet packet of sections
energy-moment
neural network
feature extraction
