【作者】 王宏；

【导师】 赵建；

【作者基本信息】 西安电子科技大学 ， 测试计量技术及仪器， 2007， 硕士

【摘要】 本课题属于自选预研。故障诊断技术是近40年发展以来的以适应工程实际需要而形成的多学科交叉的一门综合学科。随着现代科学技术的发展,现代航空装备的结构越来越复杂,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。一般电子电路的故障诊断可分为模拟电路故障诊断和数字电路故障诊断。数字技术的广泛应用和高速发展,使得数字电路的故障诊断研究取得了空前的发展。对于模拟电路由于其元件具有容差并存在非线性等原因,使得模拟电路的故障诊断较数字电路的故障诊断复杂得多,其发展比较缓慢,应用也不够广泛。但是,在一个完整的系统中,数字电路并不能完全取代模拟电路,所以对模拟电路故障诊断的研究已成为新的研究热点。故障字典法作为模拟电路故障诊断方法中最典型最具有实用价值的方法,本课题针对具有容差的电路,将经典的故障字典法与模式识别中的理论方法相结合,对字典法中的模糊集划分,节点优选等关键技术进行了研究,将贝叶斯决策应用到对节点状态模糊集的划分当中,将故障编码法应用到节点优选以及故障隔离中,完成字典的编制,实验表明,该字典提高了故障字典对具有容差的电路硬故障的诊断能力。最后本文对如何在保证诊断精度的前提下减少测前工作的计算量,提出了新的见解,使故障字典在实际应用中具有更好的可操作性。更多还原

【Abstract】 Testing and diagnosis on circuits and electronic devices are important fordevelopment andmaintenanceofsafeandreliableelectronicsystems.Althoughtest anddiagnosis on digital circuits and systems have been developed to a high level, test anddiagnosis on analog and mixed-signal circuits and systems still are evolved veryslowlydue to the lack of simple fault models and the presence of component tolerances andcircuit nonlinearities。According to collected statistics, although most part of anelectronic system is digital, about 80% of the faults occur in the analog segment. Sofaultdiagnosisofanalogcircuitismoreimportantnowadays.A further research to the fault dictionary method is presented in this paper andseveral techniques are applied in order to enhance the fault dictionary such as:application of Bayesian Theory, application of integer coding when building dictionaryand optimum test points selection. According to the experiments on circuit withtolerances, it shows that the application of these improves the capability of dictionaryonfaultdetectionandisolationwhenusingtodealwiththecircuitwithtolerance.Although these techniques improve the accuracy and definition of faultdictionary, workload increasing badly, this brings lots of work to simulation before test,this paper also makes improvement for example: using Bayesian Theory alternativelybases on the practical value (such as mean value and sigma of each fault we get bysimulating before test ). Noting that the circuit scale becomes more complex, faultdiagnosis requires heavy computations and large data structures. An artificial neuralnetworkwillbestudiedinfurtherresearch.更多还原