机构地区: 电子科技大学自动化工程学院
出 处: 《计算机辅助设计与图形学学报》 2010年第5期852-857,共6页
摘 要: 针对非线性模拟电路故障的复杂性和非线性,提出一种基于多重分形消除趋势波动分析和支持向量机的故障诊断方法.应用多重分形消除趋势波动分析方法处理信号,提取能够精细表征信号能量分布的奇异状况和几何特征分布概率的多重分形特征;将提取的多重分形特征作为支持向量机的输入样本,利用支持向量机的分类功能对电路工作模式作出故障决策.通过对Duffing混沌电路的模拟仿真实验,证明了该方法的有效性. A fault diagnosis method based on multi-fractal detrend fluctuation analysis(MF-DFA) and support vector machine was proposed,aiming at complex and nolinear in nolinerar analog circuit faults.Two content are researched which include the multifractal detrend fluctuation analysis and fault diagnosis based on multifractal spectrum features.Firstly,output voltage signals from the test node are obtained and the fault feature vectors are extracted by multifractal detrend fluctuation analysis method.Then the features are input into the one-against-all multiclass SVM to identify the different fault cases.Fault diagnosis simulation experimental results on Duffing chaos circuit shows the effetiveness of the proposed method.
关 键 词: 多重分形分析 消除趋势波动分析 非线性模拟电路 故障诊断
领 域: [电子电信]