基于数据挖掘的电路故障分析
发布时间：2018-12-10 16:48
【摘要】：随着科技发展,自动化技术水平的不断提高,小到人们日常生活,大到航空、航天、军事国防,电子设备的使用日趋广泛,其安全性、可靠性受到越来越多的关注。而电子设备的可靠性与其自身的电路系统直接相关,进行电路的故障分析一方面可以发现电路系统潜在的故障模式,另一方面可以对故障电路进行故障元器件的定位,是提高电子设备可靠性和安全性的一个重要方法。此外,系统客观的电路故障分析可以在一定程度上指导电路系统的设计,改善产品制作工艺,具有十分重要的研究意义。本文进行电路故障分析的主要思路是通过对电路系统进行正常仿真和潜在故障仿真,获取仿真数据,使用数据挖掘中的二分K均值聚类算法分析数据,解决了电路系统中潜在故障模式的划分问题;之后采用最近邻分类算法和基于规则的分类算法解决了电路系统的潜在故障分类模型构建问题;以此为基础,提出一种基于分类模型和多探测点的故障元器件定位方法,解决了故障电路中故障元器件的定位问题;本文的主要内容如下。首先介绍了数据挖掘的基本技术,包括数据挖掘的基本流程,数据集的构建,数据的预处理以及常见的分类方法和聚类方法;此外,对电路故障分析的一些概念和分析方法从数字电路、模拟电路和数模混合电路的角度分别进行了介绍。然后本文介绍了电路系统的正常仿真和故障仿真方法,具体内容包括电路激励设计方法,正常元器件和故障元器件的建模方法,电路预处理,潜在故障注入方法,仿真数据的存储;并应用介绍的方法完成了示例电路的仿真。接下来对仿真结果进行了数据提取和参数化处理,形成规范化的数据集,采用改进的二分K均值聚类算法对其进行无监督分类,完成了潜在故障模式的划分,以此为基础结合经验知识,形成分类算法训练集,并应用最近邻分类算法和基于规则的分类算法实现了电路潜在故障分类模型的构建,结合示例电路对提出的多探测点故障器件定位方法进行了应用,验证了该方法在故障器件精确定位方面具有较好的性能。本文的最后对提出的基于数据挖掘的电路故障分析方法进行了软件实现,该软件可以在一定程度上提高故障分析人员的效率。
[Abstract]:With the development of science and technology, the level of automation technology has been improved, and the safety and reliability have been paid more and more attention to, from people's daily life, to aviation, aerospace, military defense, electronic equipment. The reliability of electronic equipment is directly related to its own circuit system. On the one hand, the potential fault mode of circuit system can be found by fault analysis, on the other hand, fault components can be located on the fault circuit. It is an important method to improve the reliability and safety of electronic equipment. In addition, objective circuit fault analysis of the system can guide the design of the circuit system to a certain extent, and improve the manufacturing process of the product, which has a very important research significance. The main idea of circuit fault analysis in this paper is to obtain the simulation data through the normal simulation and potential fault simulation of the circuit system, and to use the binary K-means clustering algorithm in data mining to analyze the data. The problem of potential fault mode partition in circuit system is solved. Then the nearest neighbor classification algorithm and the rule-based classification algorithm are used to solve the problem of constructing the potential fault classification model of the circuit system. Based on this, a fault component location method based on classification model and multiple detection points is proposed, which solves the problem of fault component location in fault circuit. The main contents of this paper are as follows. Firstly, the basic technology of data mining is introduced, including the basic flow of data mining, the construction of data set, the preprocessing of data, the common classification methods and clustering methods. In addition, some concepts and analysis methods of circuit fault analysis are introduced from the point of view of digital circuit, analog circuit and digital-analog mixed circuit respectively. Then this paper introduces the normal simulation and fault simulation methods of circuit system, including circuit excitation design method, modeling method of normal component and fault component, circuit preprocessing, potential fault injection method. Storage of simulation data; The simulation of the example circuit is completed by using the introduced method. Then, the simulation results are extracted and parameterized to form a standardized data set. The improved binary K-means clustering algorithm is used to classify the simulation results unsupervised, and the classification of potential fault patterns is completed. Based on this, the training set of classification algorithm is formed by combining empirical knowledge, and the nearest neighbor classification algorithm and rule-based classification algorithm are used to construct the potential fault classification model of circuit. The application of the proposed multi-probe fault device location method in combination with an example circuit shows that the proposed method has good performance in the accurate location of the fault device. At the end of this paper, the proposed circuit fault analysis method based on data mining is implemented by software, which can improve the efficiency of fault analysts to a certain extent.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP311.13;TM13
本文编号：2370901
[Abstract]:With the development of science and technology, the level of automation technology has been improved, and the safety and reliability have been paid more and more attention to, from people's daily life, to aviation, aerospace, military defense, electronic equipment. The reliability of electronic equipment is directly related to its own circuit system. On the one hand, the potential fault mode of circuit system can be found by fault analysis, on the other hand, fault components can be located on the fault circuit. It is an important method to improve the reliability and safety of electronic equipment. In addition, objective circuit fault analysis of the system can guide the design of the circuit system to a certain extent, and improve the manufacturing process of the product, which has a very important research significance. The main idea of circuit fault analysis in this paper is to obtain the simulation data through the normal simulation and potential fault simulation of the circuit system, and to use the binary K-means clustering algorithm in data mining to analyze the data. The problem of potential fault mode partition in circuit system is solved. Then the nearest neighbor classification algorithm and the rule-based classification algorithm are used to solve the problem of constructing the potential fault classification model of the circuit system. Based on this, a fault component location method based on classification model and multiple detection points is proposed, which solves the problem of fault component location in fault circuit. The main contents of this paper are as follows. Firstly, the basic technology of data mining is introduced, including the basic flow of data mining, the construction of data set, the preprocessing of data, the common classification methods and clustering methods. In addition, some concepts and analysis methods of circuit fault analysis are introduced from the point of view of digital circuit, analog circuit and digital-analog mixed circuit respectively. Then this paper introduces the normal simulation and fault simulation methods of circuit system, including circuit excitation design method, modeling method of normal component and fault component, circuit preprocessing, potential fault injection method. Storage of simulation data; The simulation of the example circuit is completed by using the introduced method. Then, the simulation results are extracted and parameterized to form a standardized data set. The improved binary K-means clustering algorithm is used to classify the simulation results unsupervised, and the classification of potential fault patterns is completed. Based on this, the training set of classification algorithm is formed by combining empirical knowledge, and the nearest neighbor classification algorithm and rule-based classification algorithm are used to construct the potential fault classification model of circuit. The application of the proposed multi-probe fault device location method in combination with an example circuit shows that the proposed method has good performance in the accurate location of the fault device. At the end of this paper, the proposed circuit fault analysis method based on data mining is implemented by software, which can improve the efficiency of fault analysts to a certain extent.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP311.13;TM13
【参考文献】
相关期刊论文 前2条
1 崔江;王友仁;;一种新颖的基于混合故障字典方法的模拟故障诊断策略[J];电工技术学报;2013年04期
2 董晨皓，林争辉，秦建业;模拟电路故障诊断的辨识法及其方程[J];上海交通大学学报;1995年01期
相关博士学位论文 前1条
1 徐晨曦;基于解轨迹多项式分解的非线性电路故障诊断研究[D];电子科技大学;2009年
相关硕士学位论文 前3条
1 张菲菲;基于DES理论的数模混合电路的可测试性与故障诊断研究[D];合肥工业大学;2006年
2 祝丽华;基于磁场映像的电路板故障诊断研究与实现[D];中北大学;2010年
3 沈旭;基于FPGA的数字电路故障诊断系统设计及关键技术研究[D];南京航空航天大学;2014年
，本文编号：2370901
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2370901.html
教材专著
