基于可测性分析和支持向量机的模拟电路故障诊断

Fault diagnosis method for analog circuit based on testability analysis and support vector machine

针对低可测性模拟电路的测试问题,提出了一种模拟电路故障诊断的新方法。该方法首先计算被测电路的可测性,利用可测性提供的信息对被测电路进行模糊组划分,组成可诊断的元件集,并引入支持向量机完成对故障的分类识别。可测性分析理论上确定了被测电路可诊断的元件集,支持向量机结构简单,泛化能力强,以模拟和混合信号测试标准电路的实验结果证明了基于可测性分析和支持向量机的模拟电路故障诊断方法是有效的,其故障诊断率大于99%。

A new method for diagnosing fault in analog circuit with low testability is presented in this paper. The testability of circuit under test （CUT） is analyzed. According to the information provided by testability, the ambiguity group is determined and an optimal set of testable components is obtained. Support vector machine （SVM） is introduced to identify the faults of CUT. Testability theoretically determines all of the diagnosable components of CUT. SVM has the advantages of simple structure and strong generalization ability. Experimental results for analog and mix-signal benchmark circuits demonstrate the efficiency of the proposed method for diagnosing analog circuit fault, which is based on testability analysis and SVM. The fault diagnosis accuracy is better than 99%.