基于独立成分分析与支持向量机的电路测试方法

发布时间：2018-06-27 21:35

本文选题：模拟电路 + 特征提取　；参考：《湖南师范大学》2015年硕士论文

【摘要】：在高度发达的工业化进程中,为了满足安全性和环境的要求,对电路的测试要求也日益增高。随着国际间的竞争越来越激烈,电路的则试技术也日益发展,如何对电路测试的数据进行有效提取特征值和分类是目前工业发展中最关心的问题。当前基于独立成分分析(Independent component analysis,ICA)与支持向量机(support vector machine,SVM)勺故障诊断测试方法可以有效的解决上述问题。它的优点在于不需要有精确的先验知识,也不需要去构建复杂的数学模型,仅仅利用在线或者离线的数据,对数据之间的那种关系进行挖掘,最终获取电路的工作状态,完成系统的模拟故障诊断和检测。独立成分分析方法属于数据驱动的故障诊断和检测的方法之一,是一种基于多变量的高阶统计过程控制方法,它可以更有效的提取特征值;支持向量机可以更好的解决小样本、高维数、非线性等问题。因此独立成分分析方法结合支持向量机技术在电路测试、模拟故障诊断等领域都有重大的应用价值。本文的主要研究工作如下：1、针对模拟故障诊断过程中的数据具有非高斯性的特点,深入研究了自适应核函数的独立成分分析。对自适应核函数的独立成分分析从理论和实践上做了详细的分析和描述,并且对比了传统的基于nfomax ICA和主元分析(PCA)在实验的精度和所发时间进行验证,得出自适应独立成分分析方法是有效可行的。2、针对单一的高斯核函数是局部函数,学习能力强但泛化性比较差;多层感知机核函数是全局函数,泛化性强但学习能力比较弱。考虑系统不同场合下的需求,本文将结合两类核函数的优点,考虑模拟电路测试信号特征,构造自适应核函数。对自适应的独立成分分析方法处理后的数据,用自适应核函数支持向量机的方法进行分类,能更好的解决高维数、小样本、非线性等问题,最后本文用高斯核函数和多层感知机核函数去处理模拟电路故障,且从样本训练时间和测试精度上进行对比。3、把自适应的非高斯独立成分分析及支持向量机方法的具体诊断步骤进行详细说明,并且用S alley-key的带通滤波和差动放大电路以及手机测试中所产生的数据验证了该方法的可行性。
[Abstract]:In the process of highly developed industrialization, in order to meet the requirements of safety and environment, the requirements of circuit testing are also increasing day by day. With the international competition becoming more and more fierce, the trial technology of circuit is developing day by day. How to extract the characteristic value and classify the data of circuit test effectively is the most concerned problem in the current industrial development. The current fault diagnosis and testing methods based on Independent component Analysis (ICA) and support Vector Machine (SVM) can effectively solve the above problems. Its advantage is that it does not need to have accurate prior knowledge, does not need to build a complex mathematical model, only use online or offline data to mine the relationship between the data, and finally obtain the working state of the circuit. Complete system simulation fault diagnosis and detection. Independent component Analysis (ICA), one of the data-driven fault diagnosis and detection methods, is a multivariable based high-order statistical process control method, which can extract eigenvalues more effectively, and support vector machine can better solve small samples. Problems such as high dimension, nonlinearity, etc. Therefore, independent component analysis (ICA) combined with support vector machine (SVM) technology has great application value in circuit testing, analog fault diagnosis and so on. The main work of this paper is as follows: 1. The independent component analysis (ICA) of adaptive kernel function is studied in detail in view of the non-Gao Si property of the data in the process of simulating fault diagnosis. The independent component analysis of adaptive kernel function is analyzed and described in detail in theory and practice, and the accuracy and time of experiment based on nfomax and principal component analysis are compared. It is concluded that the adaptive independent component analysis method is effective and feasible. For a single Gao Si kernel function is a local function, the learning ability is strong but the generalization is poor, and the multi-layer perceptron kernel function is a global function with strong generalization but weak learning ability. Considering the requirements of the system in different situations, this paper will combine the advantages of two kinds of kernel functions, consider the characteristics of analog circuit test signals, and construct adaptive kernel functions. For the data processed by the adaptive independent component analysis (ICA) method, the method of adaptive kernel support vector machine can be used to classify the data, which can better solve the problems of high dimension, small sample, nonlinearity and so on. Finally, the Gao Si kernel function and the multi-layer perceptron kernel function are used to deal with the analog circuit fault. And compared with the sample training time and test accuracy, the adaptive non-Gao Si independent component analysis and the specific diagnostic steps of support vector machine method are explained in detail. The feasibility of the method is verified by using the bandpass filter and differential amplifier of S alley-key and the data generated in the mobile phone test.
【学位授予单位】：湖南师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN710;TP18

【相似文献】