模拟电路故障最优特征搜索与识别  

Optimal fault feature search and recognition in analog circuit

作　　者：田成来[1] 彭敏放[1] 宋丽伟[1] 谭虎[1] 沈美娥[2] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082 [2]北京信息科技大学,北京100101

出　　处：《仪器仪表学报》2012年第7期1549-1555,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60673084;60973032;61173108);湖南省自然科学基金(10JJ2045)资助项目

摘　　要：为解决模拟电路故障诊断中故障特征选取的难题并提高故障识别率,提出一种故障最优特征矢量候选集的搜索算法,并构造了模式最佳邻居查询规则来选取故障模式最优特征进行故障识别。以故障信号小波包分解的频段能量值构造故障样本的初始特征矢量,搜索识别多种模式的特征子矢量生成最优特征矢量候选集,查询模式最佳邻居确定其最大邻域,在综合判据监督下选取故障最优特征完成故障识别。模拟诊断实例表明,选取的最优特征在诊断模拟电路故障时具有满意的准确率。In order to solve the difficulty of fault feature selection and improve the fault recognition rate in analog circuit fault diagnosis, a search algorithm for the candidate sets of optimal fault feature vectors is proposed, and the query rules of pattern best neighbors are constructed to select the optimal fault features and recognize faults. Firstly, the band energy values of wavelet packet decomposition of fault signals are taken as the preliminary feature vectors of fault samples. Secondly, the adequate feature sub-vectors identifying many fault patterns are searched to construct the candidate sets of optimal feature vectors. Thirdly, the pattern best neighbors are queried to determine their largest neighborhood in feature space. Finally, the optimal fault features are selected according to the comprehensive criterions to complete the pattern recognition of fault samples. Analog circuit diagnosis examples show that the selected optimal features can recognize fault samples with satisfied classification accuracy.

关 键 词：故障诊断 模拟电路 最优特征搜索 模式最佳邻居 小波包分解 

分 类 号：TM131[电气工程—电工理论与新技术]