贝叶斯分类器在汽车发动机喷射故障诊断中的应用研究重庆大学硕士学位论文学生姓名:郑宏指导教师:向宏教授专业:计算机软件与理论(软件工程领域)学科门类:工学重庆大学软件学院二O一一年十一月ResearchonEjectionFaultDiagnosisofAutomobileEngineBasedonBayesianClassifiedforWaveformAnalysisAThesisSubmittedtoChongqingUniversityInPartialFulfillmentoftheRequirementfortheDegreeofMasterofSoftwareEngineeringByZhengHongSupervisedbyProf.XiangHongMajor:ComputerSoftwareandTheory(SoftwareEngineering)CollegeofSoftwareEngineeringofChongqingUniversity,Chongqing,ChinaNovember2011重庆大学硕士学位论文中文摘要I摘要喷射故障一直是困扰汽车发动机的一大难题,如何快速准确的对故障原因进行诊断,是新时代环境下的发展要求。在汽车发动机喷射故障诊断中,其诊断方法一直停留在传统的阈值判断的基础上,公开的资料中很少涉及到深入讨论故障诊断的算法实现。本文基于波形分析的方法,通过贝叶斯分类器来实现车载系统的喷射故障智能诊断。该故障诊断系统的研究主要包括以下几个方面的工作:①分析了贝叶斯理论所应用的领域,探索了智能算法与传统的专家系统的差异,并针对贝叶斯理论应用于汽车发动机喷射故障诊断领域进行了可行性分析;②根据汽车发动机喷射过程的波形特点,从已有的实验数据中分析特征提取规则,研究与之适合的特征选择方案;③编写了汽车发动机故障诊断系统管理平台,以用于功能模块的扩展性管理,业务管理和样本数据管理。该系统包括三个部分:写入固件的贝叶斯分类器(即智能诊断算法模型),网络版和离线版的诊断管理系统平台。通过对已有的诊断系统进行深度剖析,结合具体的需求,采用VisualStudio平台,使用C语言和VB进行混合开发。在该系统中,改进了贝叶斯分类器学习样本的存储方案,通过与专家系统的结合优化特征向量的选取,从而提高了模型判断的正确率。实例表明,基于波形分析的贝叶斯分类器模型能够较好地应用于汽车发动机的喷射故障诊断。与传统的专家系统或者单纯考虑贝叶斯理论的方法相比较,在诊断性能上有明显的改善,具备工程应用的价值。目前该诊断系统已初步尝试应于中国汽车研究院汽车发动机故障诊断系统平台中,运行情况稳定良好,诊断错误率低,实现了人工智能在汽车发动机故障诊断领域的应用,在一定程度上减少了人工干预,使诊断结果更加的客观可靠,更重要的是原有的系统并未因为引入的算法复杂性而导致性能的明显下降,达到了预期目标。关键词:喷射故障诊断,汽车发动机,波形分析,贝叶斯分类器重庆大学硕士学位论文英文摘要IIABSTRACTInjectionfaultshavebeentroubledcarenginesforalongtime.Thereisanewrequirementthathowtodiagnosisfastandaccurate.Intheejectionfaultdiagnosisfieldsofautomobileengine,olddiagnosemethodsarestillbasedonthetraditionaldeterminationofthresholdandfewofpublicpapersinvolvedsolvingtheuncertainproblems.ThepapercontrivesarapiddiagnosemethodbasedonwaveformanalysisviabuildingaspecialBayesianClassified.Inordertogetthebettercorrectrate,thismodelimprovedthestoresolutionofleaningsamplesforBayesianClassifiedbyusinganexpertsystemtooptimizetheselectionsoffeaturevectors.Theexampleshowsthatthismodelcouldbewellforrapidejectionfaultdiagnosis,solveuncertainproblemsandraisecorrectrate.Comparedwithexpertsystemandsuchonlyfocusontheimprovementoftheoryitself,thismodelhasabetterresultandthevalueofengineering.Thesystemhasachievedtheresultsinthefollowingareas:1.AnalysistheappliedfieldoftheBayesiantheory,explorethedifferencesbetweenthetraditionalexpertsystemandtheintelligentalgorithmandconductafeasibilityanalysisforcarjetenginefaultdiagnosisbasedonBayesiantheory;2.Accordingtothewaveformcharacteristicsofautomotiveengineinjectionprocess,thispaperanalysisfeatureextractionrulesfromavailableexperimentaldataforresearchingthefeatureoptionprograms;3.Writeasuiteofsoftwaremanagementsystemofcarenginefaultdiagnosisplatformtomanagetheexpansionmodules,businessesandthesampledata.Thesystemconsistsofthreeparts:Bayesianclassifierwhichwritinginthefirmware(i.e.intelligentdiagnosisalgorithmmodel),onlineandofflineversionofthediagnosticmanagementsystemplatform.Inthispaper,thereisadeepanalysistoexistingvehiclediagnosticsystemofCAERI.SystemisbasedonVisualStudioplatform,mixeddevelopedbyCandVBprogramlanguage.Atpresent,thediagnosticmodelhasbeeninitiallytrytobeintheautomobileenginefaultdiagnosissystemplatformofChinaAutomotiveEngineeringResearchInstitute,stableandgoodoperation,diagnosticerrorrate,toachievetheartificialintelligenceenginefaultdiagnosisinautomotiveapplications,toacertainextent,reducehumanintervention,sothatmoreobjectiveandreliablediagnosticresults,moreimportantly,becausetheintroductionoftheoriginalsystemisnotaresultofthe重庆大学硕士学位论文英文摘要IIIcomplexityofthealgorithmperformancedecreasedsignificantly.Thediagnosticsystemimprovesthefunctionalityoftheoriginalplatformandthediagnosticaccuracy,breakingtheoldplatformtorelyheavilyonmanualinterventiontocompletetodeterminelimitations.Keywords:Ejectionfaultdiagnosis;Automobileengine;Waveformanalysis;Bayesianclassified重庆大学硕士学位论文目录IV目录中文摘要..........................................................................................................................................I英文摘要........................................................................................................................................II1绪论.........................................................................................................................................11.1研究背景...................................................................................................................................11.2国内外现状...............................................................................................................................21.3论文研究目的及主要内容.......................................................................................................41.4论文的组织结构.......................................................................................................................52贝叶斯理论及相关技术...................................................................................................62.1贝叶斯方法...............................................................................................................................62.2贝叶斯决策.......................................................................