外文翻譯--關(guān)于北歐的疲勞實(shí)驗(yàn)室的比較—測(cè)量結(jié)果不確定值的反映.doc

3附錄附錄1英文原文Reflectionsregardinguncertaintyofmeasurement,ontheresultsofaNordicfatiguetestinterlaboratorycomparisonMagnusHolmgren,ThomasSvensson,ErlandJohnson,KlasJohanssonAbstractThispaperpresentstheexperiencesofcalculationandreportinguncertaintyofmeasurementinfatiguetesting.SixNordiclaboratoriesperformedfatiguetestsonsteelspecimens.Thelaboratoriesalsoreportedtheirresultsconcerninguncertaintyofmeasurementandhowtheycalculatedit.Theresultsshowlargedifferencesinthewaytheuncertaintiesofmeasurementwerecalculatedandreported.Nolaboratoryincludedthemostsignificantuncertaintysource,bendingstress(duetomisalignmentofthetestingmachine,“incorrect”specimensand/orincorrectlymountedspecimens),whencalculatingtheuncertaintyofmeasurement.SeverallaboratoriesdidnotcalculatetheuncertaintyofmeasurementinaccordancewiththeGuidetotheExpressionofUncertaintyinMeasurement(GUM)1.Keyword：Uncertaintyofmeasurement,Calculation,Report,Fatiguetest,LaboratoryintercomparisonDefinitions：RStressratioFmin/FmaxFForce(nektons)AandBFatiguestrengthparameterssandSStress(megapascals)NNumberofcycles.IntroductionThecorrectorbestmethodofcalculatingandreportinguncertaintyofmeasurementintestinghasbeenthesubjectofdiscussionformanyyears.TheissuebecameevenmorerelevantinconnectionwiththeintroductionofISOstandards,e.g.ISO170252.Thediscussion,aswellasimplementationoftheuncertaintyofmeasurementconcept,hasoftenbeenconcentratedonwhichequationtouseoronadministrativehandlingoftheissue.Therehasbeenlessinterestinthetechnicalproblemandhowtohandleuncertaintyofmeasurementintheactualexperimentalsituation,andhowtolearnfromtheuncertaintyofmeasurementcalculationwhenimprovingtheexperimentaltechnique.Onereasonforthismaybethattheaccreditationbodieshaveconcentratedontheveryexistenceofuncertaintyofmeasurementcalculationsforanaccreditedtestmethod,insteadofonwhetherthecalculationsareperformedinasoundtechnicalway.Thepresentinvestigationemphasizestheneedforamoretechnicalfocus.Onetestingareawhereitisdifficulttodouncertaintyofmeasurementcalculationsisfatiguetesting.However,thereisguidanceonhowtoperformsuchcalculations,e.g.inRefs.3,4.Toinvestigatehowuncertaintyofmeasurementcalculationsareperformedforfatiguetestsinreallife,UTMIS(theSwedishfatiguenetwork)startedaninterlaboratorycomparisonwhereone4ofthemostessentialpartswastocalculateandreporttheuncertaintyofmeasurementofatypicalfatiguetestthatcouldhavebeenorderedbyacustomeroftheparticipatinglaboratories.Forcostreasons,customersoftenaskforalimitednumberoftestspecimensbut,atthesametime,theyrequestalotofinformationaboutalargeportionofthepossiblestress-lifeareafromfewcycles(highstresses)tomillionsofcycles(lowstresses)andevenrun-outs.Thewaythecalculationwasmadeshouldalsobereported.Theoutcomeconcerningtheuncertaintyofmeasurementfromtheprojectisreportedinthisarticle.ParticipantsSixNordiclaboratoriesparticipatedintheinterlaboratorycomparison:oneindustriallaboratory,tworesearchinstitutes,twouniversitylaboratoriesandonelaboratoryinaconsultancycompany.Twoofthelaboratoriesareaccreditedforfatiguetesting,andathirdlaboratoryisaccreditedforothertests.Eachparticipantwasrandomlyassignedanumberbetween1and6,andthisnotificationwillbeusedintherestofthispaper.ExperimentalprocedureTheparticipantsreceivedinformationaboutthetestspecimens(withoutmaterialdata),togetherwithinstructionsonthewaytoperformthetestandhowtoreporttheresults.Theinstructionswerethattestsshouldbeperformedasconstantloadamplitudetests,withR=0.1atthreedifferentstresslevels,460,430and400Map,withfourspecimensateachstresslevel,atatestfrequencybetween10and30Hz,witharun-outlimitat6510cyclesandinanormallaboratoryclimate(0203Cand5015%relativehumidity).Thiswasconsideredasatypicalcustomerorderedtest.Thetestresultsweretobeusedtocalculateestimatesofthetwofatiguestrengthparameters,AandB,accordingtolinearregressionofthelogsandlongvariables,i.e.loglogABN.ThereportedresultshouldincludeboththeestimatedparametersAandBandtheuncertaintiesinthemduetomeasurementerrors.Thereportshouldalsoincludetheconsiderationsandcalculationsbehindtheresults,especiallythoseconcerninguncertaintyofmeasurement.Severalpropertiesweretobereportedforeachspecimen.Themostimportantonewasthenumberofcyclesuntilfractureorifthespecimenwasarun-out(i.e.survivedfor6510cycles).ThetestsweretobeperformedinaccordancewithASTME-466965andISO5725-26.ASTME-466-96doesnottakeuncertaintyofmeasurementintoaccount；However,ASTME-466-96mentionsthatthebendingstressintroducedowingtomisalignmentmustnotexceed5%ofthegreateroftherange,maximumorminimumstresses.Therearealsorequirementsfortheaccuracyofthedimensionalmeasurementofthetestspecimen.Allparticipantsusedhydraulictestingmachines.Thetestspecimensweremadeofsteel(yieldstress375390Map,andtensilestrength670690Map,tabulatedvalues).Thetestspecimensweredistributedtotheparticipantsbytheorganizer.Results5TheprimarylaboratoryresultsthatshouldbecomparedaretheestimatedWhalercurves.Inordertopresentallresultsinthesameway,theorganizertransformedsomeoftheresults.TheWhalercurvesreportedbytheparticipantsareshowninFig.1.Itcanbeseenthatthereareconsiderabledifferencesbetweenlaboratories.Anapproximatestatisticaltestshowsasignificantlaboratoryeffect.MaterialscatteralonecannotexplainthedifferencesintheWhalercurves.Inordertoinvestigateifthelaboratoryeffectwassolelycausedbythemodelinguncertainty,weestimatednewparametersfromtherawdatawithacommonalgorithm.Wethenchosetouseonlythefailedspecimensandtomaketheminimizationinthelogarithmiclifedirection.TheresultsareshowninFig.2.Aformalstatisticalsignificancetestwasthenmade,andtheresultofsuchatestshowsthatthedifferencesbetweenthelaboratoriesshowninFig.1couldbeattributedonlytomodeling.UncertaintyofmeasurementcalculationsOneofthemostimportantobjectiveswiththisinvestigationwastocomparetheobserveddifferencesbetweenlaboratorytestresultswiththeirestimateduncertaintiesofmeasurement.Theintentionwastoanalyzetheuncertaintyanalysesassuch,andtocomparethemtothestandardprocedurerecommendedintheISOguide:GuidetotheExpressionofUncertaintyinMeasurement(GUM)1.Thelaboratoriesidentifieddifferentsourcesofuncertaintyandtreatedthemindifferentways.Thesesourcesaretheloadmeasurement,theloadcontrol,thesuperimposedbendingstressesbecauseofmisalignmentandthedimensionalmeasurements.Implicitly,laboratorytemperatureandhumidity,specimentemperatureandcorrosioneffectsarealsoconsidered.Inaddition,theresultsshowamodelingeffect.ThedifferentlaboratorytreatmentsofthesesourcesaresummarizedinTable1.SpecificcommentsonthedifferentlaboratoriesAlllaboratoriesgavetheirlaboratorytemperatureandhumidity,butdidnotconsiderthesevaluesassourcesofuncertainty,i.e.theinfluenceoftemperatureandhumiditywasneglected.Thisconclusionisreasonableforsteelinthetemperaturerangeandhumidityrangeinquestion7.Laboratory1.Theuncertaintyduetotheappliedstresswasdeterminedtakingloadcellanddimensionaluncertaintiesintoaccount.ThemathematicalevaluationwasmadeinaccordancewiththeGUM.Specimentemperaturewasmeasured,butwasimplicitlyneglected.Themodelingproblemwasmentioned,butnotconsideredasanuncertaintysource.Laboratory2.Thereportcontainsnouncertaintyevaluation.Theuncertaintiesintheloadcellandthemicrometerareconsidered,butneglectedwithreferencetothelargematerialscatter.Specimentemperaturewasmeasured.Modelingproblemsarementionedbyacommentregardingthechoiceofloadlevels.Laboratory3.Thereportcontainsnouncertaintyevaluation.However,theaccuracyofthe6machineisgivenandtheloadwascontrolledduringtheteststobewithinspecifiedlimits.Thebendingstressesweremeasuredononespecimen,buttheirinfluenceonthefatigueresultwasnottakenintoconsideration.Laboratory4.Theuncertaintiesintheloadcellandthedimensionalmeasurementsareconsideredinanevaluationofstressuncertainty.ThemethodfortheevaluationisnotinaccordancewiththeGUMmethod,butwasperformedbyaddingabsoluteerrors.Thebendingstressinfluenceandthecontrolsystemdeviationsareconsidered,butnotincludedintheuncertaintyevaluation.Thefailurecriterionismentionedandregardedasnegligible,andcorrosionismentionedasapossiblesourceofuncertainty.Laborat