摘要
为了最大限度地消除粗晶材料超声检测时,晶粒散射波对有用信号的严重干扰,提高接收信号的信噪比,将小波神经网络引入粗晶材料超声检测信号处理领域中。在训练小波神经网络时,采用了改进的梯度下降算法。该网络有一个动态的权值,它随误差变化而调整。结果表明,小波神经网络应用在粗晶材料超声检测信号的降噪时,能够达到较理想的降噪效果。
In order to minimize the disturbance of the scattering wave induced by crystal-grain-structure of the material to the useful signals and raise the signal-to-noise ratio（SNR）,wavelet neural network is introduced to the signal denoising for ultrasonic detection of coarse-grain materials.The improved gradient descent algorithm has been used to train the wavelet neural network.In the process of training,the network has a dynamic weight,which can be updated automatically according to the gradient descent with adaptive learning rate.The experimental results show that wavelet neural network is an effective method in signal denoising of ultrasonic detection.
出处
《噪声与振动控制》
CSCD
北大核心
2010年第5期145-148,共4页
Noise and Vibration Control
基金
国家自然科学基金(10774164)
南阳师范学院青年基金(QN2009023)
关键词
声学
降噪
小波神经网络
粗晶材料
超声检测
acoustic
denoising
wavelet neural network
coarse-grain materials
ultrasonic detection
