【作者】 黄璐璐；

【导师】 李训铭；

【作者基本信息】 河海大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 模拟系统故障诊断课题是一个广泛研究的前沿课题之一，模拟电路的可靠性决定了电子设备系统的可靠性，具有重要的实际意义。 本文主要研究了容差模拟电路故障诊断的BP神经网络方法和故障树分析法。BP神经网络的联想记忆功能、容错性和鲁棒性以及很好的非线性映射能力，使这一诊断方法明显优于传统的诊断方法，本文尝试了通过训练样本的改进来增强网络的泛化能力；故障树分析法源自于系统可靠性分析设计，其解释基于规则，易于实现，并且充分考虑到知识工程师的领域经验，适于实现故障诊断专家系统，针对模拟电路信号的连续性与元件的容差性，本文提出了一种适用于模拟电路的模糊故障树分析法，该方法基于电路节点信息，诸如节点电压值及其波形与电阻值等建构故障树，使用模糊推理的方法，得到各个故障事件发生的可信度，并以此排序，指导故障诊断的进行。最后，对模拟电路的BP神经网络方法和故障树分析法作了比较。更多还原

【Abstract】 The analog fault diagnosis research is one of the forefronts of the testing field, one of the reason is that the reliability of electrical systems depends on that of analog circuits and systems.This paper focuses on the analog circuit fault diagnosis with tolerances by using BP neural network and fault tree analysis. With the classical pattern recognition theory, the back propagation neural network (BPNN) are applied to analog circuit fault diagnosis. The robustness, associated memory and nonlinear mapping of BPNN make this method more advantageous over traditional methods. For improving on recognition ability of network, this paper experiments with new training samples.Fault tree analysis (FTA) originated from analysis technique for systematic reliability. The diagnosis results of FTA are easy to be interpreted in that its course is based on rule. To analog circuit with tolerances, this paper present fuzzy fault tree analysis based on node information of circuit. This method uses the node information, for example node voltage value and node resistance vale, to build fault tree, then calculates the reliability of basic event by fuzzy reasoning. Finally, This paper compares BPNN with Fuzzy FTA in analog fault diagnosis.更多还原