基于矩阵扰动分析的模拟电路故障诊断方法

Fault Diagnosis Method for Analog Circuits Based on Matrix Perturbation Analysis

为了实现模拟电路的故障检测、故障定位和参数辨识的一体化处理,便于工程实施和降低故障诊断成本,提出一种基于矩阵扰动理论的模拟电路故障诊断与参数辨识方法.首先,对故障电路的采样时间序列值进行曲线拟合,得到电路故障相位偏移信息,并作为一个故障特征;其次,将采样序列构成一个方阵,求解该方阵的迹作为另一个故障特征;第三,以相位偏移信息和响应矩阵的迹随被诊断器件参数的变化而变化的对应关系为基础,结合两个故障特征,建立故障模型;最后,通过对两个国际标准电路诊断的实验验证结果显示,该方法故障定位准确率在98.5%~100%范围内,故障参数辨识误差在-1.2%~1.72%范围内.

To integrate the fault detection, fault localization and parameter identification of analog circuits in one system and reduce the cost and facilitate the engineering implementation of fault diagnosis, an fault diagnosis and parameter identification method for analog circuits based on matrix perturbation theory was proposed. First, curve fitting for the sampled time series of the faulty circuit was conducted,and the phase deviation of the circuit was treated as one fault signature . T h e n , a matrix was built using the sampled time series, and the trace of this matrix was used as the other fault signature. Finally, the phase deviation and trace were used as joint fault signatures, and the fault diagnosis model was established according to the correspondence between the changes of the fault signatures and the fault device parameter variations. T he experimental results of two international standard circuits show that the accuracy of fault location ranges from 98.5% to 100% and the error of fault parameter identification is in the range of - 1.2 % to 1.72% .