Research Of Analog Circuits Fault Classification And Diagnosis Based On Cloud Model Theory

With the increasing scale of integrated circuits, circuit fault diagnosis has been a hottopic of concern to investigators since the1960s. We mainly focus on extraction andselection of analog circuit fault feature, classification fault diagnosis and test point selectionof testability design in this paper. Method of generating fault samples based on cloud modeltheory, concept diagnosis of analog circuit soft fault, method of test point select of faultdictionary based on fuzzy cloud group, etc theoretical and practical issues are involved.The main research work and achievements include:(1) Proposed method of generating fault samples based on cloud model theory.For sample distribution of fault samples and fault feature extraction in analogelectronic circuit fault classification, two methods about generation of analog circuits faultsamples based on cloud model theory are proposed in this paper, namely fault cloudcharacteristic-value samples and fault cloud simulation samples. Method of generating faultcloud characteristic-value samples is regard the corresponding cloud eigenvalues of theoriginal fault samples as the training samples and test samples according to the statisticalproperties of cloud model theory, thus filtered noise and avoid the error between measuredata and sample data. Method of generating fault cloud simulation samples is,reconstructing a number of new samples based on the cloud characteristic-value of theoriginal sample using the forward cloud algorithm, which is suitable when the number oftraining samples is less or it’s difficult to obtain simulation sample in particular cases.Design the structure of neural network according the two samples and train, analysis,achieves fault diagnosis finally. Experimental results showed that the two generatingmethod of faults samples had achieved good results in testing and diagnosing of analogcircuit fault.(2) Proposed a soft fault concept diagnostic model represented by qualitative conceptbased on cloud model theory.As the analog circuit fault diagnosis method cannot fully represent the correspondingfault state of the soft fault, a soft fault concept diagnostic model represented by qualitativeconcept based on cloud model theory is proposed in this paper. The diagnostic modelrepresented multiple test point voltage range with soft-divided concept using cloudtransformation method, generated atomic concept of the test point voltage represented bythe cloud model characteristic-value, and promoted it to an understanding qualitativeconcept of test point voltage, generated the concept table of fault category associated withrange of faulty components’ parameters and combined by the qualitative concept ofmultiple test point voltage, and finally completed a full description of the fault conditionbased on the formal concept and classification and diagnosis of analog circuit soft fault in accordance with this table.(3) Proposed an test point selection method of analog circuit based on fuzzy cloudgroup of fault dictionary.In test point selection method based on fault dictionary techniques, aimed at thehard-divided problem of fuzzy group division rules in building an integer coding faultdictionary, the cloud model theory such as cloud transform, similar cloud and other is usedin this paper, an test point selection method of analog circuit based on fuzzy cloud group offault dictionary is proposed. The method converted the fuzzy group divided basis which isinvolved in test point selection of integer coding fault dictionary from absolute quantitativevalue to relative qualitative concept, achieved new test point selection method whichstructure fuzzy cloud group based integers coding fault dictionary using the similar cloudalgorithm of cloud model. This method fully considered the uncertain factor such as thefuzziness and randomness of analog circuit fault condition, made up the defects the "0.7V"subjective hard-divided simultaneously of fuzzy group.