一种LMD和近似熵算法的模拟电路特征提取方法
A Method for Extracting Analog Circuit Feature with LMD and Approximate Entropy Algorithms

针对模拟电路故障信号的非线性和非平稳性，提出了用局域均值分解（Local mean decomposition，LMD）和近似熵算法对模拟电路进行特征提取的方法。利用LMD算法把电路故障信号分解为一系列乘积函数（Product functions，PF），再选取前3个PF分量，求它们的近似熵，作为故障的特征向量。电路发生不同故障时，其输出响应信号的复杂度不同，经LMD分解后的PF分量的复杂度就更不相同，而近似熵可以表征时间序列的复杂度，故用LMD加近似熵可以有效提取故障电路的信息。在对故障进行分类判别时，使用核Fisher判别分析，得出各故障的诊断精度。仿真结果显示，本文的特征提取方法在改善故障电路特征的同时提高了诊断准确度，平均分类精度为97.86%。
The local mean decomposition (LMD) and approximate entropy methods were proposed to extract features of an analog circuit in order to process the non-linear and non-stationary faulty circuit signals. These signals were firstly decomposed into the sum of product functions(PF) according the LMD algorithm. Approximate entropy features, as the feature vectors of the signals, were calculated with the first three product functions. The complexity of output response signals varies from different faulty circuits; the complexity of the PF after LMD are greatly different for the same reason. Meanwhile, the approximate entropy can represent the time complexity so that the features of the faulty circuit can be effectively extracted by combining the LMD with the approximate entropy algorithms. Finally the kernel Fisher discriminant analysis was used to classify faults and detect them accurately. The simulation results show that this feature extraction method can not only extract the features of a faulty circuit but also classify them accurately, and the average classification accuracy is 97.86%