液压泵外文报告加翻译Word格式.docx
《液压泵外文报告加翻译Word格式.docx》由会员分享,可在线阅读,更多相关《液压泵外文报告加翻译Word格式.docx(11页珍藏版)》请在冰豆网上搜索。
Experimentsforderivedfactorswithapplicationtohydraulicgearpumps
C.J.Sexton,S.M。
LewisandC。
P.Please
UniversityofSouthampton,UK
[ReceivedJune1999.RevisedOctober2000]
Summary.Forexperimentsonmechanicalproductscomposedofseveralcomponents,suchasahydraulicgearpump,conventionalmethodsofdesigningandimplementingfactorialexperimentscanbeimpracticalbecauseoftheprohibitivecostsofobtainingcertaincomponentswithfactorssettoprespecifiedvalues.Afurtherdifficultyisthatoftensomeofthefactorsthatarebelievedtoinfluencetheproduct’sperformancearenotfeaturesofasinglecomponentbutarederivedasfunctionsofthedimensionsofseveralcomponentsarisingfromtheproduct’sassembly.Experi-mentsareproposedwhichuseasampleofmeasuredcomponentstoexploretheinfluenceofsuchderivedfactors.Analgorithmicmethodforobtainingefficientdesignsispresentedandappliedtofindingplansforstudiesonthegearpump。
Anexperimentonthepumpisdescribedwhichinvolvedbothconventionalandderivedfactors。
Thisexperimentledtonewknowledgeonhowtoimprovetheengineeringdesignofthepumpand,inparticular,onhowtoimproveitsrobustnesstothevaryingpressuresthatareexperiencedinoperation.
Keywords:
Algorithm;
Components;
Factorialexperiment;
Mechanicalengineering;
Optimaldesign;
Qualityimprovement;
Robustness
1.Introduction
Thedesignofaconventionalfactorialexperimentspecifiesthelevelsofeachfactorandhowtheyshouldbecombinedtoformruns。
Whentheobjectiveoftheexperimentistoinvestigateandimprovetheperformanceofaproductassembledfromseveralseparatecomponents,e。
g。
amechanicalorelectricaldevice,twomethodsofobtainingtherequiredcombinationsoffactorlevelstorealizethedesignarecommonlyused.Oneapproachistomanufactureone-offcomponentstohavethefactorvaluesthatarerequiredfortheexperiment。
Alternatively,samplesofthevariouscomponentscanbetakenfromproductionand,throughmeasur-ingthefeaturesofinterest,componentsfoundwiththerequiredproperties。
Typically,therequiredfactorlevelsareat,ornear,theextremesofthefactorrangesobtainedinproduction,solargesamplesofcomponentsmustbemeasuredtofindsufficientfortheexperiment。
Forsomeproducts,thecostsoftheseapproachesareprohibitive.AnexampleisthehydraulicgearpumpshowninFig.1,wheremachiningtheindividualgearteethonthedriveanddrivengearstoprecisespecifications,ormeasuringtheirprofiles,isslowandcostly.Afurtherdifficultywithmechanicalproductsisthatoftenacomponentcannotbeusedinmorethanoneexperimentalproduct,e。
whenrunningandtestingtheproductproducescomponentwear,asinthegearpump.
Forsuchsituations,weproposeanapproachwhichtakesamuchsmallerrandomsampleofcomponentsfromproduction,measurestheirrelevantfeaturesandassemblesthemto
Addressforcorrespondence:
C.J.Sexton,DepartmentofMathematics,UniversityofSouthampton,Highfield,Southampton,S0171BJ,UK.
E—mail:
C。
J.Sexton@maths。
soton。
ac.uk
2001RoyalStatisticalSociety0035-9254/01/50155
formasetofprototypesfortestingthatgivesasmuchinformationaspossibleonthefactorialeffectsofinterest.Thusfullcontrolissacrificedoverthevaluestakenbyafactor,butsomeControlisretainedoverwhichfactorvaluesarecombinedintheexperiment。
Suchfactorscanberegardedassemicontrolled;
seealsoHarville(1974)。
Anaddedcomplicationformanymechanicalproductsisthatthemeaningfulfactorsarenotthemeasuredfeaturesofthecomponentsbutfactorsderivedfromthem,suchasadistancebetweentwocomponents,orotherfunctionsofseveralvariablesthatrelateinengineeringtermstotheproduct'
sperformance。
Wepresentanalgorithmicmethodofobtainingefficientdesignsforexperimentsinvolvingderivedfactors。
InSection2thehydraulicgearpumpisdescribed,andSection3definesthedesignproblem。
AnalgorithmforgeneratingdesignswhenallthefactorsarederivedisgiveninSection4;
thisalgorithmhasbeenusedtofindadesignforapilotstudyonthepump.Section5considersanexperimentwheresomefactorscanbefullycontrolledandotherscannot.Anextensionofthealgorithmisusedtoobtainadesignforthisexperiment。
InSection6theanalysisofthedatafromtheexperimentisoutlined,togetherwiththenewknowledgegainedonhowtoimprovetheengineeringdesignofthepump。
2.Thehydraulicgearpump
Hydraulicgearpumpsareusedinawidevarietyofmachines,e。
g.agriculturalandconstructionvehiclesandaeroplanes,toprovideflexiblepowerfromtheenginetoliftinggearorancillaryequipment.Thepumpunderinvestigationhereformspartofthepowersteeringmechanismofanagriculturaltractor。
AsshowninFig。
1,itconsistsofadrivegearandadrivengearwhichsitintwobearingblocks,floatingwithinahousing,withaflangeplateononesideandacoverontheothertopresstheblocksontothegears.Asthedriveshaftturns,thegearteethmeshanddrawhydraulicfluidthroughthepump.However,atendencyforsomeofthefluidtoleakbackaroundtheinternalcomponentscanleadtoareductionintheefficiencyofthepump。
Toimprovetheengineeringdesignbyreducingsuchleakage,themanufacturerwishedtoinvestigatetheinfluenceofthemanygeometricalaspectsofthecomponentsonpumpleakage,definedasthedifferencebetweenthetheoreticalandactualflowratesoffluidthroughthepump。
Becauseoftheveryhighpressuresthatareexperiencedbythehydraulicfluid,thedesignmustaccommodatetheflexingofcomponentswhichwouldotherwisecreatesubstantialgapsbetweencomponentsandresultinhighleakage.Withinthepump,theflexingispartlycontrolledbyacarefuldesignofregionsontheflangeandcoverplateswherehighpressurefluidisallowedtoenter,andtherebytobalancetheforcescreatedbythepumpinginthebearings。
Suchbalancedfloatingbearingsarecentraltotheoperationofthepump.Attemptstomodelthebehaviourofthesepumpsbyusingmathematicalmodelshavenotbeenverysuccessful,primarilybecauseofthestrongrelationshipbetweenfluidflowandflexingmovementandthedynamicnatureofthegearmotionwithintheoil.Consequentlyexperimentsonasoftwaremodel(e。
g.Aslettetal。
(1998))arenotanoptionforthistypeofstudy.Realprototypesmustbeusedwhichincurcostsofmanufacture,measurementandtesting,andleadtorestrictionsonthesizeofanexperiment。
Thefinalpartofthemanufacturingprocessinvolvesrunningthepumpsothatthegearscutintothebearingblocks,creatingasealagainsttheleakageofoil.Thisbedding-inprocesscreatesanadditionalrestrictionasitchangesfeaturesofthecomponentsofacompletedpumpsothatitscomponentscannotbereusedinanotherprototype。
Forexample,whenatestedpumpisdisassembled,abrasionandpittingcanbeseenonthewallsofthebearingblocks。
Asafirstattempttoexperimentonthepump,asmallpilotstudywasplannedtoestablishlaboratorytestingprocedures,andtogainsomeinsightsintohowtheleakagemightarise。
Inthisstudyonlyfeaturesrelatedtothedrivegear,thedrivengearandthetwobearingblocks(knowncollectivelyasthegearpack)werevaried,andalltheotherfeatureswereheldconstant.Asampleofcomponents,sufficienttomake12pumps,wasavailable。
Therelevantdimensionsofthesecomponentswerecarefullymeasured,andadesignwasthenneededtospecifywhichcomponentsshouldbeselectedandassembledtomakeeachofthe12pumps.Althoughthespecificreasonsforleakagearenotknown,theengineersidentifiedthreepossibleleakagepathsthroughthepump.Foreachpathaderivedfactor(side-gap,clearanceandgearform)wasdefinedgivingthesizeofthepathasanexplicitformulaintermsofthegeometricaldimensionsofthecomponents。
Theclearanceisbasedonthedifferencebetweenthediameterofthedriveshaftandthatofthejournalbearingintowhichitfits,thegearformisafunctionoftheprofileofthegearsandtheside-gapcorrespondstothesizeofthegapsbetweenthegearsandbearingblocks.Anaimofthepilotstudywastoobtaininformationonwhichofthesederivedfactorsmightbeimportantforleakagesothatmoredetailedfollow-upexperimentscouldbefocusedontheappropriateleakagepaths。
3。
Thedesignproblem
Todemonstratetheimportanceofsolvingthisproblem,Table1givestwoofthemanypossibleplansforthepilotstudyon12gearpacks。
Herethevaluesofeachderivedfactorarescaledsothattheirminimumandmaximumpossiblevalues,calculatedacrossallvaluesthatcanbederivedfromallpossibleassembliesofthecomponents,are一1and1respectively.Thesedesignsarerealizablefromthesamesetsofmeasuredcomponentsthroughdifferentchoicesofthespecificcomponentsforassemblyintoeachofthe12gearpacks。
Fig.2showseachofthedesignsasasetofpointswithinthethree-dimensionaldesignspaceforthederivedfactors。
Thecurrentexperimentalpracticeistoassembleprototypesatrandom.Thedisadvantagesofthisapproachareillustratedbydesign1ofTable1whichoffersverylittleinformationonhowvaryingthesizeofclearanceaffectstheleakage;
11ofthe12clearancesliebetween0。
23and0.30,andsevenofthesehavevalue0。
27.Incontrast,design2exploresthisfactoracrossamuchwiderrangeofpracticalimportance,namelyfrom-0。
49to0。
91。
Fig.