摘要
损伤结构的动力特性具有局部时变的特征,小波变换在时域和频域都具有表征信号局部特征的能力,小波包分析利用可以伸缩和平移的可变视窗能够聚焦到信号的任意细节,因此可以对损伤结构的非线性动力特性能进行有效的分析。提出运用小波分析提取结构损伤特征向量的方法和基本原理,并进一步用神经网络进行损伤位置和程度的检测。文章通过一个两层框架的模型对小波神经网络和传统的BP网络的损伤识别精度作了对比。研究表明,小波神经网络的抗噪声能力较强,损伤识别的效果更好,运用小波神经网络进行结构损伤识别精度要优于传统的BP网络。
Dynamical characteristic of the damage structure is local time - variant. Wavelet transform can signify the detail of signal in time region and frequency region. The wavelet packet technique can focus on each point of signal with alterable window. So it can make effectual analyse to nonlinear dynamical characteristic of the damage structural. The method of extracting damage is proposed in this paper, and using neural network for structure damag elgenvector e detection by wavelet transform It is compared with the traditional BP network in damage identification accuracy by a tow layers frame mode. The study indicates that it has advantages of the noiseproof ability and the effectiveness of damage identification. The identification accuracy of the wavelet neural network is better than the traditional BP network.
出处
《噪声与振动控制》
CSCD
北大核心
2006年第6期43-45,共3页
Noise and Vibration Control
基金
国家自然科学基金项目(编号:60275004)
关键词
振动与波
小波变换
损伤检测
小波神经网络
结构
vibration and wave
wavelet transform
damage detection
wavelet neural network
structure
作者简介
王步宇（1968-），男，河北省抚宁县人，硕士，工程师，主要从事工程结构的损伤检测和健康监测等。
