【作者】 芦湘冬；

【导师】 何怡刚；

【作者基本信息】 湖南大学 ， 电工理论与新技术， 2004， 硕士

【摘要】 模拟电路故障诊断经过三十多年发展，已经有很多故障诊断理论和方法，但是这些理论和方法只适合于非容差或小容差线性电路。随着电子技术的发展，特别是超大规模模拟电路和模数混合电路发展，对模拟电路故障诊断提出了新的挑战，需要探索新的模拟电路故障诊断理论与方法。人工神经网络用于模拟电路故障诊断是一门崭新的学科，近年来，受到了普遍的关注和重视。数据融合作为许多传统学科和新兴工程领域相结合而产生的一个新的前沿学科，超越了在军事领域的应用，已经在许多领域都得到了广泛的应用。 虽然模拟电路故障诊断、人工神经网络和数据融合技术在各自的领域或者两两相互结合的领域取得了一定的发展，但是目前关于人工神经网络和数据融合技术相结合对模拟电路进行故障诊断的文献还未有所见。本文把数据融合技术引入人工神经网络来对模拟电路进行故障诊断，形成一种新的方法。 本文在传统的模拟电路故障字典法的基础上引入了数据融合中的数据挖掘技术，优化了故障字典的建立。本文在神经网络对模拟电路进行故障诊断的基础上提出了基于人工神经网络的数据融合方法，主要目的是为了解决故障诊断中的容差问题并提高诊断的正确率和诊断速度。本文探讨了RBF神经网络结合D-S证据理论的D-S证据神经网络在模拟电路故障诊断上的应用。神经网络的联想记忆功能、容错性和鲁棒性及非线性映射能力和数据融合技术对多源数据的分析、综合、推理能力相结合，使得这一方法明显优于传统的神经网络诊断方法。通过对电路进行仿真表明，利用所提出的证据神经网络方法对模拟电路故障进行了较准确的诊断，达到了较好的目标。更多还原

【Abstract】 In the past thirty years, it has developed a great deal of theory and approaches for fault diagnosis. But these methods only adapt to linear circuits with non-tolerance or mini-tolerance.Along with development of electrical technology,especially VLSI and mixed circuits, It brings up new challenge to fault diagnosis of analog circuits.It is in great need of new theory and approaches to it.It is a new subject for fault diagnosis to make use of Artifical Neural Network. In recent years, researchers paid more attention to ANN than before. The application of ANN has turned into a succeed aspect for ANN. Data fusion has been used widely in many areas besides military area as a new advanced knowledge which was combined with many traditional konwledge and new engineering areas.Though fault detect and diagnosis for analog circuits, ANN and data fusion has made great progress in each area and some areas combined with each two of them, there are fewpapers in fault detect and diagnosis for analog circuits with method combined with Artificial Neural Network and data fusion. In this paper, data fusion was imported in ANN to diagnosis fault of analog circuits. It comes to be a new technology.In the paper, we used data mining in data fusion on the basis of traditional DC dictionary approach to optimize the course of construction a fault dictionary. The paper present a data fusion technology with ANN based on diagnosis for analog circuits with ANN. It aimed to deal with parameter tolerance, correctness and diagnosis speed. A Radial Basis Function(RBF) ANN using Dempster-Shafer theory of evidence which is abbreviated evidence network is presented to diagnose analog circuits. The combination of association memory function, robustness, nonlinear mapping ability of ANN and analysis, synthesis and reasoning ability to multisource data of data fusion made it better than traditional approaches. The simulation to analog circuits can indicate that it is a good approach to diagnose analog circuits correctly. It proved to be feasible under lots of experiment.更多还原