摘要
提出了一种基于神经网络的扬声器故障检测方法.通过扫频仪激励扬声器,采集扬声器的响应信号,利用小波包分解的方法时频分析响应信号,得到各频段的能量;对分解后各频段信号的特征能量进行提取,规一化特征能量;把规一化后的特征能量作为人工神经网络的输入,通过BP神经网络对扬声器故障进行分类识别.实验对165个扬声器进行识别,识别率为95.8%.实验结果表明,该方法简便有效,具有实用价值.
A fault diagnosis method based on artificial neural networks was presented.After the loudspeaker was incentived by the sweeping LAU, the response signal of loudspeaker can be acquired. The response signal was decomposed by means of wavelet packet decomposition of time-frequency analysis method, each band decomposition characteristics energy was extracted, the feature energy was normalized put as the artificial neural network input, then the loudspeakers faults were classified by BP Neural networks. Experiments on 165 loudspeakers were designed.The identification rate was 95.8%. Results show that this method is effective and feasible.
出处
《天津科技大学学报》
CAS
2008年第1期46-48,62,共4页
Journal of Tianjin University of Science & Technology
基金
天津市科技发展计划资助项目(06YFGPGX08900)
