【作者】 冯志红；

【导师】 王炜； 林志贵；

【作者基本信息】 天津工业大学 ， 信息与通信工程， 2008， 硕士

【摘要】 随着电子技术的飞速发展电子设备的组成和结构变得越来越复杂,规模也越来越庞大,为了提高系统的安全性和可靠性,故障诊断技术越来越受到人们的重视,成为各界研究的热点方向之一。另外,电子系统中容易出现故障的部分往往是模拟电路,因此,从某种意义上说模拟电路的可靠性决定了电子设备系统的可靠性,所以对模拟电路故障诊断的研究已成为故障诊断的重中之重。本文首先采用两种基于信息融合的故障诊断方法——神经网络结合模糊融合和神经网络结合D-S证据理论融合两种方法,其基本思想是神经网络对测试信息进行预处理,融合方法对预处理结果进行综合诊断。通过两种方法对一个具体的模拟电路进行仿真试验,结果表明:D-S证据理论融合方法在处理模拟电路故障诊断中的不确定性方面优于模糊融合。其次,对D-S证据理论进行改进并进行仿真,结果表明:改进的D-S证据理论在处理证据间的冲突问题方面明显优于基本的D-S证据理论,因此提高了诊断结果的可靠度。最后,基于支持向量机的信息融合方法进行模拟电路故障诊断,实验表明:支持向量机特别适用于小样本情况下的模拟电路故障诊断,该方法具有很多优点,在模拟电路故障诊断中具有很好的应用前景。为了提高诊断的准确性和效率,本文以MATLAB7.0和Access2003为平台设计了一个基于信息融合的电路故障诊断系统。更多还原

【Abstract】 As electronic technology develops rapidly the compose and structure of electronic equipments gets more and more complicated, scale gets larger, the fault diagnosis technology is regarded by more people to improve system’s safety and security and becomes one of being researched hotspot aspects. Besides, analog circuits fault more easily in electron systems, so in a way the analog circuits’ security decides the security of electronic equipments, therefore it becomes more important to research analog circuits’ fault diagnosis.In this thesis, firstly, two fault diagnosis methods based on information fusion are adopted, neural network combines fuzzy method and neural network combines D-S method, the essential idea is that primary dispose of mesurement information is done by neural network and integrated diagnosis of disposed result is done by fusion methods. A specific analog circuit’s experiment is done by two methods, results indicate that the later is more predominant than the former in the uncertainty of disposing the analog circuits fault diagnosis. Secondly, D-S is proved and simulation results indicate that improved D-S is more predominant than the basic D-S in disposing the conflictions among evidences and diagnosis result is more reliable. Finally, a method is adopted based on SVM’s information fusion, experiments indicate that it is more applicable for SVM to diagnose the analog circuits’ faults with very small samples, SVM has many merits and a better application foreground.To improve the accuracy and efficiency of diagnosis, a circuit fault diagnosis system based on information fusion is designed with the MATLAB7.0 and Access2003 software platform.更多还原