搜索
114()Vol.11No.420108JOURNALOFAIRFORCEENGINEERINGUNIVERSITY(NATURALSCIENCEEDITION)Aug.201031,2,2,2(1.,710089;2.,710038):,,,,,8,,:;;;;DOI:1013969/j1issn11009-3516120101041002:V241101:A:1009-3516(2010)04-0006-05,[1],;[2],,,,,[3-5][5],,,,,,,,1[6]13:2010-04-05:(513270501):(1964-),,,,;E-mail:hyt666@tom.com(1966-),,,,,.1Fig.1Influencingfactorssystemofairborneproducts2(SupportVectorMachine,SVM)[7-9],,,,211SVMT={(x1,y1),(x2,y2),,(xn,yn)}(XY)n,xiRn,yiRXYP(x,y),c(x,y,f),R(f)=c(x,y,f)dP(x,y)f(x,)=(x)+b,(0)T,Remp(,b)=1nni=1|yi-(x)-b|,(1),Remp(,b):,b,i,3i:F(i,3i)=ni=13i+ni=1is.t.[(xi)+b]-y+iyi-[(xi)+b]+3i,3i0,i=1,2,,n(1):L(,3,,3,,C3,,3)=ni=1(3i+i)-ni=1i[yi-(xi)-b++i]-ni=13i[(xi)+b-yi++3i]-C32[cn-()]-ni=1(3i3i+ii)(2),b,3i,C30,3i0,i0,3i0i0,f(xi)=(xi)+byi,,,,:|f(xi)-yi|-,:minW(,3)=12ni,j=1(i-3i)(aj-a3j)(xixj)+ni=1(i+3i)+ni=1yi(i-3i)s.t.ni=1(i-3i)=00i,3iC(3):74:f(x)=ni=1(-i+3i)(xixj)+b(4),,:f(x)=ni=1(-i+3i)K(xi,xj)+b(5)K(xi,xj)212SVMX(),F(X,),n(X1,1),(X2,2),,(Xn,n),f(X,)f(X,0),[10]S,X,^,,,^39,1,9(I),,8600hI,28100h,10-6/hC138,1Tab.1InfluencingfactorsofairborneproductsofeachcityABCDEFGHIC1011017011110011018011011011C2012011013011019011014016011C3017011013011015011110014013C4012014013017017015017016017C5013017014110011018011012013C6014012015013019012017018016C7011013011019011016011011011C8012012014013016012018015014C9017012013016017012018013013C10015015014015015015015015015C11015015013015015015015015015C12015015013015015015015015015C13015015014015015015015015015,,,01001,22100h(:10-6/h)Tab.2Statisticalfailurerateofproductevery100hofeachunit(10-6/h)/hABCDEFGH10015020080120330150400260200200200100180380180440300300250300100180500250630470400330400170270600180670550500440480340450800360880800600600640510620980540113010502SVMFig.2FlowchartforfailurerateforecastingbasedonSVM8()20108,,3I33100hTab.3Forecastedproductfailurerateevery100h/h100200300400500600/(10-6h-1)272192312189439139500182673182863186(t)(4),R(t),(t)R(t):R(t)=exp-t0(t)dt(6):TMBF=0R(t)dt(7)TMBF,(t)(6),R(t)(7)TMBF5961h,I,3100hFig.3Testcontrastoffailureratesevery100h4Fig.4Forecastingcurveoffailurerate4,,13,,,,,,,,,:[1].GJB198915-94.[S].:,1994.CommissionofTechnologyandIndustryforNationalDefence.GJB198915-94,ParameterSelectionandIndexDeterminingRequirementofMaterielReliabilityandMaintenance[S].Beijing:ChinaStandardCommittee,1994.(inChinese)[2],,.[J].:,2002,3(1):22-25.LIUJincheng,FENGJinfu,CUIGong.AStudyofCalendarTimeControlofMilitaryAirborneEquipment[J].JournalofAirForceEngineeringUniversity:NaturalScienceEdition,2002,3(1):22-25.(inChinese)94:[3].[D].:,2008.LIUBaogang.StudyonCalendarLifeConfirmingandExtendingMethodsofAirborneProducts[D].Beijing:BeijingUniversi2tyofAeronauticsandAstronautics,2008.(inChinese)[4],.11[J].,2005,34:18-23.CHENShengbin,HAOZongmin.TheResearchaboutMaintenanceInnovationandUse-lifeExtensionofAboardEquipmentforZ11Helicopter[J].HelicopterTechnique,2005,34:18-23.(inChinese)[5].[J].,1995,16(5):44-49.JIAOJingtang.StudyonChinaAviationAirborneEquipmentLife[J].ActaAeronauticaetAstronauticaSinica,1995,16(5):44-49.(inChinese)[6],,.[J].,2005,2(6):28-33.ZHANGLunwu,HEXinzhou,YANGWanjun.AnalysisoftheEnvironmentalEffectsandEnvironmentalEngineeringStandardforMateriel[J].MetalFormingTechnology,2005,2(6):28-33.(inChinese)[7]VapnikV.TheNaturalofStatisticalTheory[M].NewYork:SpringerPress,1995.[8]GunnS.SupportVectorMachinesforClassificationandRegression[R].ImageSpeechIntelligentSystemResearchGroupISIS-1-98,1998.[9],,.[M].:,2004.LIYinghong,WEIXunkai,LIUJianxun.EngineeringApplicationsofSupportVectorMachines[M].Beijing:WeaponIndus2tryPress,2004.(inChinese)[10].[D].:,2007.WUTao.StudyonAssessmentofAirForceAviationMaintenance[D].Xian:AirForceEngineeringUniversity,2007.(inChinese)(:)ResearchonSVM-basedPredictingMethodonCalendarLifeExtensionofAirborneProductsLIZheng-qi1,HEYu-ting2,SHAOQing2,WEIPeng2(1.ChinaFlightTestInstitute,Xian710089,China;2.EngineeringInstitute,AirForceEngineeringUniversity,Xian710038,China)Abstract:Inthispaperaninfluencingfactorsystemofairborneproductsisestablishedbasedonacomprehensiveconsiderationoffactorslikegeography,climate,servicesituationandmaintenancelevel.SVMregressionanalysismodelandforecastingmodelofairborneproductsbasedonSVMareintroduced.ThenalifepredictionmodelbasedonSVMisbuiltaccordingtothefailureratedataofanairborneproductequippedinsomerepresentativeconditions.Itcanpredicttheservicelifeoftheproductbyfindingouttherelationshipbetweenthefailurerateandtheinfluen2cingfactors,accordinglyitcanmakeupthelocalizationofoutfielddatastatisticalmethod.Acaseshowninthispa2perpresentsthelifeforecastingprocessofbefore-mentionedmethodandtheresultindicatesthatthismethodisofsomevalueforpracticalapplication.Keywords:airborneproducts;calendarlife;lifeextension;supportvectormachine;regressionanalysismodel01()2010