He Wei, He Yigang, Li Bing, Zhang Chaolong
School of Electrical Engineering and Automation, Hefei University of Technology, Hefei 230009, China.
School of Electrical Engineering, Wuhan University, Wuhan 430072, China.
Entropy (Basel). 2018 Aug 14;20(8):604. doi: 10.3390/e20080604.
In this paper, a novel method with cross-wavelet singular entropy (XWSE)-based feature extractor and support vector machine (SVM) is proposed for analog circuit fault diagnosis. Primarily, cross-wavelet transform (XWT), which possesses a good capability to restrain the environment noise, is applied to transform the fault signal into time-frequency spectra (TFS). Then, a simple segmentation method is utilized to decompose the TFS into several blocks. We employ the singular value decomposition (SVD) to analysis the blocks, then Tsallis entropy of each block is obtained to construct the original features. Subsequently, the features are imported into parametric t-distributed stochastic neighbor embedding (t-SNE) for dimension reduction to yield the discriminative and concise fault characteristics. Finally, the fault characteristics are entered into SVM classifier to locate circuits' defects that the free parameters of SVM are determined by quantum-behaved particle swarm optimization (QPSO). Simulation results show the proposed approach is with superior diagnostic performance than other existing methods.
本文提出了一种基于交叉小波奇异熵（XWSE）特征提取器和支持向量机（SVM）的模拟电路故障诊断新方法。首先，应用具有良好抑制环境噪声能力的交叉小波变换（XWT）将故障信号转换为时频谱（TFS）。然后，利用一种简单的分割方法将TFS分解为几个块。我们采用奇异值分解（SVD）对这些块进行分析，然后获得每个块的Tsallis熵以构建原始特征。随后，将这些特征导入参数化t分布随机邻域嵌入（t-SNE）进行降维，以产生具有判别性和简洁性的故障特征。最后，将故障特征输入到SVM分类器中以定位电路缺陷，SVM的自由参数由量子行为粒子群优化（QPSO）确定。仿真结果表明，所提出的方法具有比其他现有方法更优越的诊断性能。
