Numerical Analysis of the Finite Element Methods f

整理文档很辛苦,赏杯茶钱您下走!

免费阅读已结束,点击下载阅读编辑剩下 ...

阅读已结束,您可以下载文档离线阅读编辑

资源描述

NUMERICALANALYSISOFFINITEELEMENTMETHODSFORMISCIBLEDISPLACEMENTSINPOROUSMEDIASandraM.C.MaltaAbimaelF.D.LoulaLaboratorioNacionaldeComputac~aoCientcaRuaLauroMuller,455,Botafogo,RiodeJaneiro,RJCEP22290-160,Brazil.Finiteelementmethodsareusedtosolveanonlinearsystemofpartialdierentialequationswhichmodelsincompressiblemiscibledisplacementofoneuidbyanotherinporousmedia.Fromabackwardnitedierencediscretizationwedeneasequentiallyimplicittime-steppingalgorithmwhichuncouplesthesystem.TheGalerkinmethodisemployedtoapproximatethepressure,andimprovementvelocityeldapproximationsarecalculatedviaapost-processingtechniquewhichinvolvestheconservationofthemassandDarcy’slaw.Astabilizedniteelement(SUPG)methodisappliedtotheconvection-diusionequationdeliveringaccuratesolutions.Quasi-optimalordererrorestimatesareobtainedundersuitableregularityhypotheses.1.IntroductionWestudyniteelementapproximationsforasystemofnonlinearpartialdierentialequationsgov-erningincompressiblemiscibledisplacementintwodimensionalporousmedia.ThemathematicalmodelconsistsofanellipticsystemcomingfromtheconservationofmassandDarcy’slawandadegenerateparabolicequationexpressingtheconservationoftheinjecteduid(concentrationequation).LetbeaboundeddomainintheplaneR2withsmoothboundary@andT0axednumber.Ourdierentialsystem,underappropriatephysicalassumptions,isgivenby[1]divu=fin(0;T);(1:1)1u=K(x)(c)rpin(0;T);(1:2)withtheboundaryconditionu=0on@(0;T);(1:3)and@c@t+div(cu)div(Drc)=^cfin(0;T);(1:4)withtheboundaryandinitialconditionsDrc=0on@(0;T);(1:5)c(x;0)=c0(x)on;(1:6)wherepanduarethepressureandDarcy’svelocityofthemixture,=(x)andK(x),theporosityandpermeabilityofthemedium,respectively,=(1;2)denotestheexteriornormalto@,fthesourceandsinktermsand^c=^c(x;t)istheinjectedconcentrationatinjectionwellsandtheresidentconcentrationatproductionwells.Thediusion-dispersiontensorDwillbeconsideredasin[1],i.e.,D=D(u)=mI+jujflE(u)+tE?(u)g;(1:7)withE(u)=1juj2uu;E?(u)=IE(u);(1:8)wherem,landtare,respectively,moleculardiusion,longitudinalandtransversedispersioncoecients.Normallydispersionisphysicallymoreimportantthanthemoleculardiusion;also,lisusuallyconsiderablylargerthant,andweshallmakethisassumptioninouranalysis.Sincep(x;t)isdetermineduptoanarbitraryadditiveconstant,wenormalizeitbyimposingtheconditionZp(x;t)dx=0;t2(0;T):(1:9)Finally,wenotethatinEquation(1.2)=(c)isthelocalviscosityofthemixturewhichdependsontheconcentrationc.Suchdependenceisveryimportantintheemergenceofviscousngering,2displacementeciency,andultimateoilrecovery.Inthepetroleumengineeringliteraturethemostcommonlyusedformtorepresentis(c)=(0)h1c+M14ci4;c2[0;1];(1:10)whereM=(0)=(1)isthemobilityratio.Inspiteofbeingc(x;t)themostimportantvariableintothesolutionofsystem(1.1)-(1.6)wehavedemonstratedin[2,3,4,5]thatthevelocityeldu(x;t)hasstronglyinuenceontheaccuracyoftheconcentration.Thisfactismainlynotedforadversemobilityratiocases,i.e.,whenM1inEquation(1.4).WhenM=1,whichcorrespondstothetracerinjectionprocessesinoilreservoirs,wepresentedin[6]anumericalanalysisoftheuncoupledsystem(1.1)-(1.6)consideringanon-standardGalerkinmethodfortheconcentrationequationwiththevelocityeldu(x;t)givenbyapost-processingtechnique.Hereweshallextendthatproceduretothetwophaseincompressiblemiscibledisplacement,inotherwords,weshallassumeM1throughoutthispaper.Theorganizationofthepaperfollows.Insection2,afterintroducingthebasicnotationandas-sumptions,acontinuous-spaceapproximateproblemisdenedbyusingasequentiallyimplicittimediscretization.Insection3,consideringpressuregivenbytheusualGalerkinmethod,weproposedapost-processingtechniquetorecoveranaccuratevelocityeld.Next,thepost-processingap-proachiscombinedwithatechniqueofsubtractionofsingularitytotreatthecaseofpointsourcesandsinks.Insection4theellipticsubsystemisassumedtobesolvedandastabilizedniteelementformulation,SUPG(StreamlineUpwindPetrov-Galerkin),isusedtoapproximatetheconcentra-tionequation.Errorestimateswillbeexhibiteddemonstratingquasi-optimalconvergenceratesinL1(L2)-norm.Section5summarizesourndings.2.Preliminaries2.1NotationsandDenitionsLetm0beanarbitraryintegerand1q1,wedenotebyWm;q()theusualSobolevspace3oncontainingfunctionsvhavingnitenormkvkm;q:=XjjmZj@vjqdx1q;where=(1;2;:::;n)isamulti-index,and@:=@1x1@2x2:::@nxn.Whenq=2thenWm;2()correspondstotheHilbertspaceHm()withnormkkm;2=kkm.ThecaseH0()=L2()hasitsnormdenotedbykk0=kk,andtheinnerproduct(v;w):=Zwvdx:Wealsorequirespacethatincorporatetimedependence.LetXanormedlinearspaceconsistingofasetofunctionsdenedon,suchasmentionedabove.Ifwisafunctiondenedon[0;T]wesayw2Lp([0;T];X),1p1,iffort2[0;T],w(;t)2XandkwkX2Lp([0;T]),andwedenekwkLp([0;T];X):=ZT0kw(t)kpXdt1p:2.2RegularityoftheDataInitiallyweimposeverystrongregularityhypothesesintheanalysisofsystem(1.1)-(1.3),someofthesewillbeweakenedinsubsection3.3.Therefore,thefollowingregularityassumptionswillbemadeonfunctionsK,fandonthesolutionfu;pgof(1.1)-(1.3):(1)ThereexistconstantsKandKsuchthat,forallx2,0KK(x

1 / 31
下载文档,编辑使用

©2015-2020 m.777doc.com 三七文档.

备案号:鲁ICP备2024069028号-1 客服联系 QQ:2149211541

×
保存成功