【作者】 李云红；

【导师】 谭阳红； 汤先云；

【作者基本信息】 湖南大学 ， 电气工程， 2008， 硕士

【摘要】 模拟电路故障诊断一直以来都是十分必要且有意义的,已成为热门的研究课题。传统的模拟电路故障诊断的方法也有很多种,但是它们一般都是用于诊断开路、短路这种硬故障的,难以发现在电路中的各个元器件存在类似器件缺陷或缓慢失效之类的软故障。此外,在很多情况下,一些电路只有一个可测试点,也就是它的输出端,传统的方法根本无法对它们进行有效诊断。而基于神经网络的模拟电路故障诊断方法可以很好地解决这些问题。已用在模拟电路故障诊断中的神经网络主要有BP网络、SOFM网络,以及结合模糊理论的模糊神经网络,结合小波分析的小波神经网络,但是仍然存在问题,例如模糊度如何准确地定量化,对故障信号进行小波变换之后怎样构造能表征故障类别的特征等,都有待进一步研究[1]。量子神经网络(Quantum Neural Network)简称QNN,因其隐层神经元采用多层激励函数使网络具有了一种固有的模糊性,它能将决策的不确定性数据合理地分配到各模式中,从而减少模式识别的不确定度,提高模式识别的准确性。量子神经网络已成功应用于图像处理、气象预测及语音识别等,但用在模拟电路故障诊断中不多见,本文结合不同的故障提取方法,提出了基于量子神经网络的容差模拟电路的软故障诊断,通过仿真实验,将量子神经网络成功应用到模拟电路的软故障诊断中,并与BP神经网络相比,提高了故障诊断的正确性。更多还原

【Abstract】 Analog circuit fault diagnosis has been very necessary and meaningful, and it has become a hot topic for research. There are many ways for diagnosing traditional analog circuit fault, but they are mainly used to diagnose hardware faults such as open, short circuits, those soft faults such as similar device defects or gradually invalidation existing in various parts of the circuit are hard to be found. In addition, in many cases, there is only one testing point for some circuits, that is, its output, traditional methods are unable to effectively carry out their diagnosis.The method of analog circuit fault diagnosis based on neural network can solve these problems properly. The neural network used in analog circuit fault diagnosis are BP network, SOFM network, fuzzy neural network combined with fuzzy theory, wavelet neural network combined with wavelet analysis. But there are still problems, such as how to accurately quantify the ambiguity, how to construct features that can show types of the faults after the wavelet transform of fault signal. All of these need further study [1].Quantum Neural Network, which is simplified as QNN, has a kind of inherent fuzziness because its hidden neurons adopt multi-layer stimulating functions. It can reasonably distribute uncertainty data to every model, Thereby the uncertainty of Pattern Recognition is reduced and the accuracy of pattern recognition is improved. Quantum neural network has been successfully applied to image processing, weather forecasts and voice recognition, but using in analog circuit fault diagnosis is rare. By combining different faults extraction methods, the paper proposes soft fault diagnosis of analog circuit based on quantum neural network, applies quantum neural network successfully to the soft fault diagnosis of analog circuits through the simulation. By comparing with BP neural network, the accuracy of fault diagnosis is improved.更多还原