【作者】 黄亮；

【导师】 侯建军；

【作者基本信息】 北京交通大学 ， 交通信息工程及控制， 2012， 博士

【摘要】 模拟电路故障诊断对电子系统的设计生产和使用维护起着至关重要的作用,但是由于模拟电路的自身难点,使模拟电路故障诊断成为困扰电路发展的瓶颈,模拟电路故障诊断研究成为电路研究领域十分迫切的重要课题。为了增加模拟电路故障诊断技术的理论储备,促进模拟电路故障诊断应用技术的发展,本文结合国内外现有的模拟电路故障诊断方法,对模拟电路故障诊断的理论原理与诊断技术进行了系统深入的研究。针对模拟电路故障诊断的故障样本采集与处理、故障诊断的测试点优选、故障诊断方法及其与故障诊断相关的数学算法,提出了几种新的模拟电路系统分析与故障诊断方法,全文的主要工作与研究成果如下。1.针对模拟电路的灰色系统特性,将基于灰色关联分析与灰熵分析的灰熵关联算法应用于模拟电路故障诊断领域,计算灰熵关联度以量化待诊断电路与故障状态之间的关联程度,仅需要少量样本和测试点即可准确完成模拟电路故障诊断。提出了基于灰熵关联算法的模拟电路测试点优选方法,使用灰熵关联度量化模拟电路的测试点与故障元件之间的关联程度,为选择模拟电路故障诊断的最佳测试点提供了数学模型,客观明确的描述了测试点在模拟电路故障诊断中的优劣。2.提出了基于聚类算法与马氏距离算法的模拟电路故障诊断法,首先使用层次聚类算法或FCM聚类算法将模拟电路故障样本按照故障状态进行分类,根据各聚类中心建立故障字典,继而使用马氏距离算法完成模拟电路故障诊断。将聚类算法与马氏距离算法相结合进行模拟电路故障诊断,既能发挥聚类算法优秀的分类能力,也能发挥马氏距离算法优秀的距离计算能力,拓展了传统的故障字典法,诊断流程清晰流畅,诊断方法具有很好的实用性。3.提出将模糊隶属函数转化为模拟电路故障度量化函数,建立了常用的故障度量化函数,提出了基于故障度量化函数与决策函数的模拟电路故障度量化算法,使用模拟电路故障度量化算法计算模拟电路的故障度指数,客观准确地评价了模拟电路的故障程度,实现了模拟电路系统的精确描述。使用基于隶属函数与决策函数的模糊算法进行模拟电路故障诊断,将待诊断电路与各电路状态的相似程度由模糊量转变为数字量,精确实现模拟电路故障诊断。4.提出了基于重分类与支持向量机相结合的模拟电路故障诊断方法,研究了两种重分类方法,重新分类存在故障类重叠的故障样本,以提高模拟电路故障诊断的正确率。根据支持向量机的分类结果验证了重分类方法对降低故障诊断误诊率的有效性。本文共6章,图41幅,表54个,参考文献115篇。更多还原

【Abstract】 Fault diagnosis of analog circuits is very important in the process of electronic system design, production, using and maintain. Because there is much trouble in analog circuits, fault diagnosis of analog circuits already have been the restriction of circuit research. The research on fault diagnosis of analog circuits is an urgent subject in the circuit research field.In order to develop fault diagnosis of analog circuits, increase the theory reserves of fault diagnosis of analog circuits, put the technology of fault diagnosis of analog circuits into use, this dissertation study theory and technology of fault diagnosis of analog circuits integrally. Theory and technology of fault diagnosis of analog circuits is summarized at home and abroad. This dissertation study every field of fault diagnosis of analog circuits including the collection and handling of fault samples, the selection of optimal test points, the method and algorithm of fault diagnosis of analog circuits. Several kinds of new method of system analysis and fault diagnosis are proposed. The major study and result of this dissertation are the following.1. Grey entropy relational algorithm which come from the combination of grey relational analysis and grey entropy analysis is proposed to analyze and diagnose analog circuits because there is the character of grey system in analog circuits. Grey entropy relational degree is used to quantify the relationship between undiagnosed circuits and fault states. With the smaller samples and test points, grey entropy relational algorithm can ensure the accuracy of fault diagnosis of analog circuits. Grey entropy relational algorithm is proposed to select the optimal test points of analog circuits. Grey entropy relational degree is used to quantify the relationship between test points and fault components, grey entropy relational algorithm can ensure the objectivity of selecting the optimal test points.2. A new method based on clustering algorithm and Mahalanobis distance algorithm is proposed to diagnose analog circuits. Fuzzy C-means (FCM) clustering algorithm and hierarchical clustering algorithm are used to analyze fault samples of analog circuits. After clustering analysis, fault samples are classified, fault dictionary is structured simultaneously. Mahalanobis distance algorithm and fault dictionary method are combined to finish fault diagnosis of analog circuits. This method can utilize the excellent classification ability of clustering algorithm and the excellent distance calculation ability of Mahal anobis distance algorithm. Traditional fault dictionary method is expanded. This method has the advantages of smooth process and practical applicability.3. Fuzzy membership function is transformed to quantify the fault degree of analog circuits. Several common fault degree quantification function are set up. Fault degree quantification algorithm based on fault degree quantification function and decision function is proposed to calculate the fault index of analog circuits. Fault index of analog circuits can evaluate the fault degree of analog circuits objectively, describe analog circuit system accurately. Fuzzy algorithm based on membership function and decision function is used to diagnose analog circuits accurately. The similarity between undiagnosed circuits fault states is quantified by Fuzzy algorithm.4. A new method based on reclassification and SVM is proposed to diagnose analog circuits. Two method of reclassification are studyed to reclassify fault samples of types overlap. The classification result of SVM prove that reclassification can increase the accuracy of fault diagnosis when there is overlap area in fault samples of different fault types.This dissertation consists of6chapters, including41figures,54tables and115references.更多还原