【作者】 王佩丽；

【导师】 彭敏放；

【作者基本信息】 湖南大学 ， 电气工程， 2010， 硕士

【摘要】 模拟电路测试和故障诊断自二十世纪六十年代以来,一直是热门研究课题,至今已取得了诸多显著的理论成果,但由于模拟电路元件的非线性、具容差及其故障现象的多样性等使得其诊断问题极其复杂,现有诊断理论与方法距实际应用尚有一定的差距。目前,小波包变换理论和支持向量机的研究与应用成为新的研究热点。小波包技术作为信息处理的有力工具,其与支持向量机等相结合,为解决模拟电路故障诊断中的诸多难题提供了可能。本文主要目的在于将模拟电路故障诊断与小波包、支持向量机方面的最新研究成果相结合,探索解决模拟电路故障诊断问题的有效途径。本文首先对模拟电路故障诊断理论的研究现状进行了综述,针对模拟电路故障诊断的特征提取和故障分类这两大问题,分别阐述了小波包分析和支持向量机的基本理论,着重研究了小波包分析和支持向量机在模拟电路故障诊断中的应用,然后应用小波包分析方法和引入一种多层动态自适应优化参数的最小二乘支持向量机方法分别对电路故障情况进行分类,并用仿真实验说明了其具体实施。为了解决模拟电路故障诊断中的特征提取困难并对模拟电路故障信号进行有效的分类,本文提出了一种基于模糊优化小波包分解的模拟电路故障特征提取算法,在此基础上,提出了结合模糊最优小波包和最小二乘支持向量机(LSSVM)的模拟电路诊断方法。该法首先对模拟电路的响应信号进行小波包分解,并引入模糊准则对其优化,得到由分类能力强的最优小波包基能量值构成的特征集,然后将特征集输入LSSVM网络,以实现对不同故障类型的识别。小波包的优化分解减小了LSSVM网络的规模,从而降低了算法复杂度,加快了网络的训练时间和分类速度。模拟诊断实例表明,此方法能快速准确地实施模拟电路的故障定位。更多还原

【Abstract】 Analog circuit fault diagnosis has been an active area since the 1960s with many significant work and methods carried out. Unfortunately, the progress of analog circuit fault diagnosis from the fundamental theory and methods to practical application has been hampered by many factors such as nonlinear effects, component tolerances, poor fault models etc. At present, the study and application of Wavelet Packet and Support Vector Machine has become the reaseach hotspot in the field of fault diagnosis. It is researched with the hope that application of Wavelet Packet and Support Vector Machine to the area of analog circuit diagnosis may achieves better results. The main purpose of this paper is to combine latest research for the Wavelet Packet and Support Vector Machine with analog circuit fault diagnosis in order to explore a new way for solving the problem of analog circuit fault diagnosis.This paper firstly gives a description for analog circuit fault diagnosis, then represents the principle of Wavelet Packet Analysis and Support Vector Machine respectively aiming at the two issues including feature extraction and fault classification of analog circuit fault diagnosis,and focuses on the application of wavelet Packet Analysis and Support Vector Machine in analog circuit fault diagnosis. Also,this paper applies the Wavelet Packet Analysis method and Least Squares Support Vector Machine method whose parameters are optimized by a method called multi-layer adaptive best-fitting parameters search respectively to clssifying the fault, so as to describe the performance of the methods by two diagnostic examples.In order to solve the difficulties in the feature extraction and classification of fault signals in analog circuits,this paper first presents a new feature extraction algorithm based on optimal Wavelet Packet combined with Fuzzy-rule.Then, a new diagnosis method combined with the feature extraction algorithm and Least Square Support Vector Machine (LSSVM) is proposed. The response signals of analog circuits are preprocessed by Wavelet Packet Transform, and the Fuzzy Rule is used to find the optimal wavelet packet coefficient of which classification capacity is better.Then, the feature set which is composed of the optimal wavelet packet energy is inputted into a LSSVM network to identify different fault case.The optimal Wavelet Packet Tranform combined with Fuzzy-rule can decrease the LSSVM network size, which is helpful to reduce algorithm complexity and accelerate learning and convergence speed. The diagnostic example illustrates this method is effective and accurate for fault location of analog circuits.更多还原