模拟集成电路故障诊断与参数辨识的代数方法

Algebraic Methodology on Fault Diagnosis and Parametric Identification for Analog Integrated Circuits

为实现模拟电路参数辨识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和参数辨识方法。该方法不同于基于数字信号处理(DSP)与人工智能的方法,从代数观点出发,以被测电路响应矩阵的本征值随被诊断器件参数的变化而改变的对应关系为基础,建立故障模型。该模型将故障检测、故障定位和参数辨识一体化处理,具有易于工程实施的优点。实验结果表明,该方法的计算时间开销小,可降低测试成本,故障定位和故障参数辨识精度高,实验结果中的最大辨识误差为2.35%。

In order to achieve the parameter identification and reduce the diagnosis cost of analog circuit, unlike the methods based on digital signal processing （DSP） or artificial intelligence, a new method based on matrix perturbation theory for analog circuit fault diagnosis was proposed to construct a theoretical framework that integrates fault detection, fault location and parameter identification based on the relationship between the device parameters and response matrix eigenvalues. By comparing the proposed model with its counterparts, the experimental results showed that the proposed model outperforms its counterparts in the respect of computational cost and fault parameter identification. In two experiments, the identification errors are less than 2.35 %.