基于多频小波分析和D-S推理的电路故障诊断

Circuit Fault Diagnosis Based on Multi-Frequency Wavelet Analysis and D-S Reasoning

对于模拟电路,提出了一种对多频响应信号特征进行信息融合的故障诊断方法。选取多种频率信号作为电路的输入,对响应进行小波分析,提取各频带信号的能量特征,经归一化处理得到与电路状态对应的故障特征矢量。待测样本作为未知故障模式与故障特征矢量进行灰色关联分析,得到待测样本与各故障模式的灰色关联度。以灰色关联度构造证据合成规则所需的基本概率赋值函数,并利用D-S证据推理进行融合诊断。实验证明,与现有方法相比,本方法测试成本低、诊断准确率高,对于轻微的软故障,也能进行有效的诊断。

A novel analog fault diagnosis method applying multi-frequency response information fusion is proposed. Several signals with different frequencies are selected as the inputs of the analog circuit. The responses are processed by wavelet analysis and normalized to be the fault mode feature vectors. The test sample is associated with the fault mode feature vectors by gray correlation degrees,which constitute the basic probability assignment function. Different frequency responses in time-frequency domain act as the different evidences,which are fused by the D-S evidence reasoning to be a consistent diagnosis conclusion. Experimental results show that the proposed method is more accurate and costs less than the methods in being. It is effective even in the slight soft fault diagnosis.