基于模式識(shí)別原理的神經(jīng)網(wǎng)絡(luò)模擬電路故障診斷方法.pdf

1、學(xué)校代號(hào)：10532學(xué)號(hào)：S03094002密級(jí)：湖南大學(xué)碩士學(xué)位論文基于模式識(shí)別原理的神經(jīng)網(wǎng)絡(luò)模擬電路故障診斷方法堂焦蟲(chóng)遺厶絲名i一型全要昱!匝絲芻盈巫整!縫：險(xiǎn)目9熬援：進(jìn)塞副熬塑!圍煎透副塑援境羞望僮；血?dú)洚?dāng)信顯至程堂醫(yī)童、業(yè)名整；魚(yú)王理：l金皇塹基丕迨塞握窯旦塑；2QQ生4目2Q旦迨塞簽姑旦期；2Q籩生旦!Q目?；邵莆m會(huì)圭顯；固直鏖塑援AbstractThefaultdiagnosisoftheanalogcircuitsis

2、anadvancedsynthesisintercrossedsubject，whichisanapplicationtechnologyabsorbingthenewtheory，technologyandmethodinothersubjectsandfieldsInthepast40years，ithasdevelopedagreatdealoftheoryandapproaches，Buttheapproachesarelimi

3、tedtodealwithfaultdiagnosisbecauseofthevarietyandcomplexityoftheanalogcircuits，especiallythelarge‘scaleanalogcircuitswiththetoleranceorthesoftfaultArtifieialNeuralNetworkS(ANN)whichhavebeenoneofthemeStactiveresearchareas

4、recentlycanbecontributedtosolveproblemsinvariouspracticalfieldsInthispaper，analogcircuitfaultdiagnosisapproachbasedonpatternrecognition，andANNisexpatiatedInthepaper，basedontheexistingliteratureresearchfoundationananalogc

5、ircuitcatastrophicfaultlocationapproachbyusingfeedforwardnetworkswithback—propagationlearningisrealizedBythisapproach，thesimulationrequirementsbeforetestarereducedbecausefewertrainingsamplesareneeded，andthefaultlocationp

6、rocessisfastThismethodisveryefficientinlocationofsinglehardfaultwitcomponenttolerancesThemeasurementspacefeatureandthegeneralcharacterizationconceptofsingleanddoublesoftfaultinlinearcircuitsarepresentedAccordingtothiscon

7、cept，alinearcircuitssoftfaultlocationapproachusingsubhiddenlayerBPNNisestablishedwithelementtolerance，anditisshownthatthisapproachissuccessfulinfaultlocationAdoublefaultfeatureextractionmethodforlinearcircuitswithelement

8、tolerancesarealsopresentedBythismethod，thedoublefaultgeneralcharacterizationcanbecalculatedbysinglefaultgeneralcharacterizationwhichcanbecalculatedbysinglefaultfeatureThismethodmakessimulationbeforetestmoresimpleAccordin

9、gtotheactualprojectrequirements，practicalBPalgorithmispresentedandrealizedonpersonalcomputerAfterfurtheroptimizationandimprovements，asubhiddenlayerBPNNalgorithmwhichsupportunlimitedunitsisrealizedThealgorithmispowerful，s