Research On Circuit Fault Annlysis Based On Data Mining

With the growing using of electronic equipment,the scale and complexity are significantly increased,especially in the area of aerospace and others,so whether the electronic equipment is reliable and safe or not becomes more and more important.Circuit failure analysis is an important method of improving the safety and reliability of electronic equipment.This method can discover the potential fault mode in the circuit and locate the failure.Therefore,the circuit failure analysis is of great significance.In this thesis,the main idea of circuit failure analysis follows as below.1)The data source for data mining is solved by circuit simulation.2)The problem of feature redundancy in the sample is solved by the feature selection,so that the clustering analysis can extract the most potential fault modes by less feature subsets and reduce the complexity of the classification model.3)The dividing of the potential fault mode in the circuit system is solved by the clustering analysis.4)The construction of the potential failure classification model of the circuit system is solved by the rule-based classification algorithm and the nearest neighbor classification algorithm.On the other hand,the FMEA analysis based on the circuit simulation will obtain the failure propagation relationship between the circuit modules and the failure relation between the components and the circuit modules.Those relations provide two new ideas for failure positioning,one is the failure location based on FMEA analysis and the other is the failure location based on failure propagation directed graphs.The former can be used to locate failure directly in the phenomenological level with the state of the alarms,the later solves the problem of failure location efficiently by filtrating the candidate failure sources and the candidate non-failure sources.the main contents of the thesis are as follows:Firstly,the basic theories of data mining and circuit failure analysis are introduced,including the relationship between different circuit fault levels,the failure propagation between the circuit modules,and several common circuit failure diagnosis methods.Secondly,the circuit simulation is introduced,and the result of the circuit simulation is processing differently on the base of the different types of signals.The thesis also introduces the knowledge of feature selection including the general selection process of feature selection,clustering analysis and the judgment criteria of clustering validity,and raises a feature selection algorithm based on clustering analysis proved that the algorithm has an advantage in the efficiency and the time complexity.Thirdly,the methods of failure positioning is researched.On the one hand,the result of two-point K clustering is used for the training set of failure location methods based on classification model including the classification model of the rule-based algorithm and the classification model of the nearest neighbor algorithm.On the other hand,according to the results of FMEA analysis,the failure location based on FMEA analysis and the failure location based on the failure propagation directed graph are proposed.The above failure positioning methods are analyzed and verified by circuit examples,and the effect of failure positioning is good.Finally,the thesis completes the software realization of the whole circuit failure analysis method,which reduces the dependence on the user experience and improves the efficiency and accuracy of the failure analysis.