摘要
多余物的材质信息对于控制避免多余物的产生具有重要的意义,以往的研究都是针对声音信号展开的,本文以加速度扰动信号为研究对象,提出了基于小波变换的航天继电器多余物材质分类方法。运用三门限检测算法,实现了扰动信号的拼接,有利于特征量的集中。采用基于频域特征的材质分类方法,定义了频谱重心,实现了金属与非金属的分类识别,在实验中确定了频谱重心的分类界限是50kHz。采用小波变换定义能量分布矢量,利用BP神经网络实现了多余物微粒确定材质的分类,并在试验中利用不小于1mg的多余物粒子验证了上述算法。实验结果表明分类的准确率分别为67.78%和76.67%。本文研究的分类方法可以推广应用于其他军用电子元器件及电子装置的多余物检测应用中。
The material information is very important for preventing from generation of the remainders. According to ever research around with sound signal, this paper proposes methods of material classification based on wavelet transform for aerospace relay using acceleration-disturbing signal. The disturbing signal is jointed together for features in focus by detection algorithm with three thresholds. Metallic and nonmetallic particles are categorized by spectrum barycenter defined in frequency domain, and the boundary of classification is determined as 50 kHz in experiments. The definite material of remainder paticles are categorized based on BP neural network by energy distribution vector, which is defined by wavelet transform. The algorithms are validated by experiment using particles whose mass exceed 1 mg, and the classification accuracy is up to 67.78% and 76.67% respectively. The classification methods present in this paper can be applied in remainder detecting for other military electronic components and electronic devices.
出处
《电工技术学报》
EI
CSCD
北大核心
2009年第5期52-59,共8页
Transactions of China Electrotechnical Society
基金
国防科工委工业技术基础科研基金(FEBG27100001)
总装军用电子元器件型谱序列科研基金(QT070920503XE0300)资助项目
