火電廠制粉系統(tǒng)故障診斷研究.pdf

1、分類號(hào)UDC窬級(jí)。#弓彳屈火摯碩士學(xué)位論文論文題目：火電廠制粉系統(tǒng)故障診斷研究論文作者：趙生光蓑親、擘#‰警雹：丁萃菁教授重慶大學(xué)職稱、工作單位：J卒青教殘重厭大掌申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)名稱：動(dòng)力機(jī)械及工程論文提交日期：2002年5月12日答辯日期：2002年5月31日學(xué)位授予單位：重慶大學(xué)授位日期：年月日答辯委員會(huì)主席：何祖威教授博導(dǎo)論文評(píng)閱人：何祖威教授博導(dǎo)王富典教授級(jí)高工2002年5月12日重塞查堂堡主堂垡堡莖垂苧塑堊ABSTRA

2、CTInthispaper，thefaultdiagnosticproblemofmiddlestoragepulverizingsystemforChongQingpowerplantisdeveloped，Bymixedknowledgerepresentation、BPneuralnetwork、expertsystembasedOilknowledgetechnologyetcareintegratedinthissystemA

3、ccordingtotheoperatingandfaultcharacteristicofthepulverizingsystem，thegeneralframeoffaultdiagnosissystemforpulverizingsystemispresentedfirstly，thensamplecompilingmodule、faultstyleupdatingmodule、conditionmonitoringanddeal

4、ingmodule、faultdiagnosisandexplainingmodule、dataqueryingandanalyzingmoduleetcareestablishedMeanwhilethissystemisprogrammedwithDELPHl50anditsinterfaceiseasyforoperating。Duringthecourseofcompilingfaultsample，multimethodofk

5、nowledgerepresentation，asproducerule、fralne、neuralnetworkrepresentationetcaresynthesizedtoorganizeandexpressthefaultsampleofpulverizingsystemBeforebeingputtedintothedatabase，thesampledataisdisposedtoensuresampledatainteg

6、rityandnoredundancyDuringthecourseofdevelopfaultdiagnosticmethod，theinfluencetothetrainingcirclenumber麗吐lnetworkstructure、learningrate、initialweightvalue＆doorvalueetcarediscussedBycomprehensiveanalysesandcomparing，thecom

7、parativelyrationalvalueisadoptedtobenetwork’SeigenvalueInadditiont0mendselfleamingrateBPneuralnetwork，oneparametercallederrorgradsamendingcoefficientisfomld，theresultindicatesthatthefaultdiagnosissystemcouldmakeerrorreac

8、htOtheaimvaluequicklyandefficaciouslyforallpulverizingsystemfaultsamplesIfthereareplentyofsamples，thissystemCandiagnosepulverizingsystemfaultexactly，andthesystemalsoincludesexplainingmechanismwhichcanprovidefaultreasonan

9、dfaultdissolvingmethodItisobviousthatthesystemCanhelptopreventpulverizingsystemfaulthappeningandensurethepulverizingsystemrunsafelyandreliablyKeywords：middlestoragepulverizingsystem，BPneuralnetwork，expertsystem，faultdiag