基于小波包能量谱和NPE的模拟电路故障诊断  

Analog circuit fault diagnosis based on wavelet packet energy spectrum and NPE

作　　者：孙健[1,2] 王成华[1] 杜庆波[2] 

机构地区：[1]南京航空航天大学电子信息工程学院,南京210016 [2]南京信息职业技术学院,南京210023

出　　处：《仪器仪表学报》2013年第9期2021-2027,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(50907033)资助项目

摘　　要：提出采用小波包能量谱和邻域保持嵌入作为预处理实行特征提取的模拟电路故障诊断方法。该方法对采集到的故障响应信号进行小波包分解,将不同频带内的能量作为故障特征值,然后利用邻域保持嵌入算法进一步提取故障特征,最后将所得到的最优故障特征输入支持向量机进行故障诊断。仿真结果表明,提出的故障特征提取方法能很好地反映故障响应信号的本质特征,不仅表现出了比其他特征提取方法更好的性能,而且最后的故障诊断中也获得了令人满意的结果。An analog circuit fault diagnosis approach utilizing wavelet packet energy spectrum and neighborhood pre- serving embedding （NPE） as preprocessing is proposed. This approach performs wavelet packet decomposition on the acquired fault response signals, and takes the energy within different frequency bands as the fault feature values, then the neighborhood preserving embedding algorithm is used to further extract fault features;finally, the obtained optimal fault features are inputted into support vector machine to identify different faults. The simulation results show that the proposed method can well reflect the essential feature of the fault response signal ; the method not only displays better performance than other methods, but also obtains satisfactory results in fault diagnosis.

关 键 词：模拟电路 故障诊断 特征提取 小波包能量谱 邻域保持嵌入 支持向量机 

分 类 号：TN707[电子电信—电路与系统]