turbulent viscosity rate limited问题

Let'stakecareofthewarningturbulentviscositylimitedtoviscosityratio****whichisnotphysical.Thisproblemismainlyduetooneofthefollowing:1)Poormeshquality(i.e.,skewness0.85forQuad/Hex,orskewness0.9forTri/Tetraelements).{whatvaluesdoyouhave?}2)Useofimproperturbulentboudaryconditions.3)Notsupplyinggoodinitialvaluesforturbulentquantities.出现这个警告，一般来讲，最可能的就是网格质量的问题，尤其是Y+值的问题；在划分网格的时候要注意，第一层网格高度非常重要，可以使用NASA的ViscousGridSpaceCalculator来计算第一层网格高度；如果这方面已经注意了，那就可能是边界条件中有关湍流量的设置问题，Yourmeshmayhavehighcelljumpsinthedomain1这个应该是湍流模型的选取与第一层网格高度之间不满足近壁处理关系而产生的问题，如果你没有使用壁面函数的话，第一层网格高度尽可能地小点儿，比如取为弦长的百万分之一左右；另外，边界条件中关于湍流量的设置不合理也会导致这个警告。2（不推荐）solve-controls-limitsMaximumTurb.ViscosityRatio加多两个0，估计是一些单元的最大Turb.ViscosityRatio超出了限定值（）恕我直言，你的这个方法只是治标不治本，他这个问题多数是由于网格尺度太大引起的。也可能是边界条件上的湍流相关参数不合理导致的。[br][br][以下内容由larky在2007年06月23日00:00am时添加][br]调大限制值可能导致发散3这是一个办法，能够解决一部分问题，有一些问题无论你怎么调整都没有用，如果出现这种情况可以通过调整初始流场，找到合适的初始值大部分能够解决，其实如果只是一开始初现这个问题，可以不作调整，除非影响到收敛性4在别的论坛上看到的：为了尽快收敛对异常的数值进行的限制，对最后收敛结果无影响1）如果边界条件设置合理，一般来说会在收敛后自动消除。2）为了加快收敛对异常的数值进行的限制（以引用2楼），是加快收敛的一种措施。3）但是如果你的问题中流场变化很大，有可能在最有还会有。4）如果网格不好会经常出现这种现象。5）如果不想看见它总是报告而影响计算速度（写屏会降低计算速度），可以在下面把它关闭：solve-control-flowlimit....具体记不住了，自己看看就知道了。5我也遇到这种情况,不过是在叠代求解的前一百多步,后面就没有了.因此我想是否是因为前面计算的误差大引起?而随着计算误差的减少,就消失了.如果是这样,就可以放心啦.6一般是边界上或是网格质量差的地方出现了奇点.由于是数值耗散,随着迭代次数越多,影响整个流场的范围越大,最终可能导致这个流场发散.如果是网格质量差的地方出现,就只能重划网格了如果是在边界上,一般是湍流相关参数设置不合理造成的,改成固定湍流比可能能解决7Whydon'tyoutryasfollows(Ifyoustillhavethesamewarning,pleasegotothenextstep.Usually,theinitailflowconditionusedfortheRSMrunisobtainedfromtheRNGk-emodelresult);Firststep:Solve-Controls-Solution-Default=iterationSecondStep:DecraseUnder-relaxationfactors=iterationThirdStep:Adaptationofcells:Iusauallyusey+andvelocitygradientconditions=iterationFourthStep:Regeneratemesh,gotostep1Ifyoursolutionstatstoconverge,youcanincreaseunder-relaxationfactors.Ifyouhaveconvergedsolutions,youcanincreasetheorderofthediscretizationparameters(forex.1st-2nd-QUICKetc.)8，onceipostedabigmessageonthisissue,iampastingthatmessageagain,youcanreadthis:{wellthisisonecommonproblemlotofpeoplehaveaskedaboutitbefore.iwilltrytosummarizetheapproachitaketosolvethisproblem.firstofall,theverybasiccauseofthiswarningisthewrongsetupofboundaryconditions.Soifyouaresurethatnothingiswrongwiththesetupoftheproblem,youcanfollowthefollowingthings.Theoriginoftheproblemliesinthefactthatifthesolvercalculatesthevalueofkandeoromega(intwoequationmodels)wrongly,itsverylikelyitwillcalculateturbulentviscositywronglyandthuswegetthewarning.Intheidealcondition,asthesolutionconvergesthewarningshouldgoawayandwealllivehappilyeverafter.Butgenerallythisdoesnothavesohappyending.Thereasonismainlywehaveacasewhichisverylargeandconvergenceisalreadydifficultandwhichisexacerbatedbythewrongcalculationsofturbulentquantities.Sowhataretheremediesforit.Theusualremedyistoswitchtocoupledsolver,andworkwithit,andthisusuallysolvestheproblem.Butmypersonalthinkingisthatifthecaseisincompressiblethecoupledsolvermaynotworkwellthere.Butyesthisisonesolution.Thesecondsolutionwhichisfarmorestableis,andifyoufailtogetthesolutionfromcoupledsolvertoo,switchtoFAS,increasethenumberofprepostiterations,makethecoarseninglevelsto4,(4ismorethanenough).Andthisconvergesalmosteveryproblem,buttherearecasewhereyoumightfailtogetconvergence.Anywayifyouarestuckwithsegregatedsolver(likeme),whataretheoptions.Firstofallifweconsiderthatthedivergenceisbecauseofturbulencequantities,wemaywanttoforcetheconvergenceonthesequantitiesbeforewemovetonextiteration.ThewayIdoisthis,Ichangethemultigridoptionsfork,etoVcycle,makethepressweepsto1postsweepsto2,andchoseBicgstabassmoother.Andletitrun.SometimesIjustwanttofirstgetthebestapproaximationofk,efortheflowfieldIhave,forthisIusuallyswitchoffthesolverformomentumequationsandjustsolvefork,eork,omegatillIgetwarningfreeturbulentfield,thenIswitchonalltheequationsandgoontoiteratefurther.Thisapproachworkswell,butithasoneproblem.ifthemeshsizeisverylargesayaround3millioncellstheneventofirstgettheturbulentquantitiestoconvergemighttakedayortwo.Sowhattodointhiscase.WheneverIhavetodocalculationsforthecasesaround2-3millioncells,Imaketwomeshesoneveryverycoarse,withsameboundaryconditionsasfinermesh(whichisofcoursearound3millioncells).NowfirstIgetconvergedsolutiononcoarsemesh,whichIcangetinhourortwo.Thengotofile-interpolate,andwritethedataforcorrespondingzones,andthenwhenyoureadthefinemeshreadthisinitialguessfromsamefile-interpolate-read.Andhereswitchoffthemomentumcalculationsdosomecalculationsonlyforturbulentquantites,(ifugetviscocitywarning,itwillsoongoaway,thoughInevergotwarningheresincethesolutionisalreadyconverged),soaftersay3-4calculationsswitchonsolverforallthequantitiesandgoontoiterate,youwillgetconvergedsolution.(welloncoarsegridyoucanuseFAStoforceconvergence,itsquitehandythere).Hopethiswillbeuseful.}