You are on page 1of 11

COMOLLOG(/LOG)

Menu

CompungDeignenivieinCOMOL
Mulphic
WalterFrei(hp://www.comol.com/log/author/walterfrei/)|Augut5,2015
(hp://www.faceook.com/harer/harer.php?
=100&p%5url%5D=hp%3A%2F%2Fwww.comol.com%2Flog%2Fcompungdeign
enivieincomolmulphic%2F) (hp://twier.com/intent/tweet?
url=hp%3A%2F%2Fwww.comol.com%2Flog%2Fcompungdeigneniviein
comolmulphic%2F) (hp://plu.google.com/hare?
url=hp%3A%2F%2Fwww.comol.com%2Flog%2Fcompungdeigneniviein
comolmulphic%2F) (hp://www.linkedin.com/hareArcle?
mini=true&url=hp%3A%2F%2Fwww.comol.com%2Flog%2Fcompungdeign
enivieincomolmulphic%2F&ro=fale)
Oneuefulutinmexperience,rareluedcapailitavailalewithinCOMOL
Mulphicitheailittocomputedeignenivie.Aumingthatouhaveaingle
ojecvefunconthaticomputedaedonourniteelementmodel,oucaneail
computehowenivethiojecvefunconiwithrepecttoanmodelinput,uingonl
thecoreCOMOLMulphicpackage.Inthilogpot,wewilllookathowtouethi
funconalit.

AnIntroductorxample:TheParallelPlateCapacitor
Lettartthiconveraonwithaverimpleprolemcompungthecapacitanceoftwo
parallelatquaremetalplate,ofidelength
,eparatedaditance
,andwithadielectricmaterialofrelavepermivit =2andwichedin
etween.Undertheaumponthatthefringingeldareinignicant(aratherevere
aumpon,utwewillueitheretogettarted),wecanwriteananalcexpreionforthe
capacitance:

where

ithepermivitoffreepace.

Wecaneaildierenatethiexpreionwithrepecttoourthreeinputtondthedeign

Wecaneaildierenatethiexpreionwithrepecttoourthreeinputtondthedeign
enivie:

NowletlookatcompungtheeameenivieuingthefunconalitofCOMOL
Mulphic.

chemacofaparallelplatecapacitormodel,neglecngthefringingeld.
Wecantartuildingamodeluingthelectrotacphicinterface.Ourdomainwille
alockoflength andheight witharelavepermivitof .Theoundarcondion
attheoomiaGroundcondion,andatthetop,anlectricPotenalcondionetthe
voltageto
.TheideofthelockhavethedefaultZeroChargeoundarcondion,
whichiequivalenttoneglecngthefringingeld.Wecanolvethimodelandndthe
voltageeldetweenthetwoplate.aedonthioluon,wecanalocalculatethetem
capacitanceaedontheintegraloftheelectricenergdenit,
,throughouttheenre
model:

Thiequaondoeaumethatoneplate(orterminal)iheldatavoltage ,whileallother
terminalinthemodelaregrounded.Theintegraloftheelectricenergdenitoverall
domainialreadcomputedviatheuiltinvariale comp1.es.intWe ,andwecanueitto
deneanexpreionforthecomputedcapacitance.Ofcoure,wewillwanttocomparethi
valueandourcomputedenivietotheanalcvalueowecandeneomevariale
fortheequane.Wecanuetheuiltindierenaonoperator,d(f(x),x),toevaluatethe
exactenivie,ahowninthecreenhotelow.

Varialeareuedtocomputethemodelcapacitance,awellatocomputetheexactcapacitance
anditenivie.Theuiltindierenaonoperator,forexampled(C_exact,L),caneuedto
evaluatetheexactenivie.
Aerolvingourmodel,wecanevaluatethecomputedtemcapacitance,compareittothe
analcvalue,andevaluatethedeignenivieanalcall.Nowletlookathowto
computetheeeniviewithCOMOLMulphic.
Theparameterthatweareconideringaectoththematerialpropereawellathe
geometricdimenionofthemodel.Whenthedeignparameteraectthegeometr,we
needtouetheDeformedGeometrinterface,whichletuevaluatetheeniviewith
repecttoageometricdeformaon.

Thedeignparameterwillaectachangeinthegeometrahown.Thehiddenfaceexperience
nodiplacementnormaltotheoundarie.

WeintroducetwonewGloalParameter,

and

,whichrepreentachangein

and

WeintroducetwonewGloalParameter,
and
,whichrepreentachangein and
.TheewilleuedintheDeformedGeometrinterface,whichhafourrelevantfeature.
Firt,aFreeDeformaonfeatureiappliedtothedomain,whichmeanthatthecomputaonal
domaincandeformaedontheappliedoundarcondion.Next,PrecriedMeh
Diplacementfeatureareappliedtotheixfaceofthedomain.Inthecreenhotelow,the
deformaon(dL)normaltothefaceiprecriedahownintheketchaove.

ThePrecriedMehDiplacementfeatureareuedtocontrolthediplacementnormaltoall
domainoundarie.
Finall,toactuallcomputetheenivie,wemutaddaenivitnodetothetud
equence.Thiihowninthecreenhotelow.Youwillwanttoentertheojecve
funconexpreion,inthicae C_computed ,awellaallofthedeignparameterthatou
areinteretedintuding.Alo,chooethevalueforthedeignparameteraroundwhichou
wanttoevaluatetheenivie.ince
and
repreentanincrementalchangeinthe
dimenion,wecanleavetheeothatzerotocomputetheeniviefor
and
.Theparametercontrollingthematerialpermivitneednopecialhandling,
otherthantochooeitaoneoftheparameterintheenivittud.
Therearetwooponintheformhownelowforthegradientmethod:

Theadjointmethodifatandoluonmewillnotdependonthenumerof
parametertudied.Infact,theadjointmethodaddonltheequivalentofa
ingleNewtoniteraontothetotalcomputaonalcot.
Theforwardmethodoluonmeincreaewiththenumerofparameter,ut
doeaddionallreturntheenivitoftheenreoluoneld,whichmae

doeaddionallreturntheenivitoftheenreoluoneld,whichmae
uefulinomecae.

Aenivitfeatureuingtheadjointmethodiaddedtothetudequence,andtheeng
howtheojecvefunconandtheparameterthatareconidered.
Aerolving,ouwillnowealetogotoReult>DerivedValue>Gloalvaluaonand
entertheexpreion fsens(dL) , fsens(dD) ,and fsens(epsilon_r) toevaluatethe
enivitofthecapacitancewithrepecttothedeignparameter.Ofcoure,oucanalo
comparetheetotheprevioulcomputedanalcenivieandoerveagreement.
Analc

Computed

88.542nF

88.542nF

354.17nF/m

354.17nF/m

1770.8nF/m

1770.8nF/m

Nowthatwehavetheaicideadownintermofcompungtheenivitofthe
capacitanceofthitem,whatelecanwedo?Certainl,wecanmoveontoomemore
complicatedgeometrie,utthereareafewpointthatweneedtokeepinmindawemove
eondthiexample.

LimitaonandcopeWhileCompungDeign
enivie
Therearetwocondionthatmutefullledforenivitanalitowork.Firt,the
ojecvefunconitelfmutedierenalewithrepecttotheoluoneld.Thimean
thatojecvefunconuchathemaximumandminimumofaeldarenotpoile.
econd,theparametermuteconnuouintherealnumerpace.Thu,integer
parameter(e.g.,thenumerofpokeonawheel)arenotpoile.
enivitcalculaonarenotcurrentlavailaleforeigenvalueprolem,norratracingor
parcletracing.
ThedeignparameterthemelvearetpicallGloalParameter,utoucanalouethe
enivitinterfacetoaddaControlVarialeFielddenedoverdomain,oundarie,edge,
andpointadeired.

Ojecvefunconaretpicalldenedintermofintegraloftheoluonoverdomainor

Ojecvefunconaretpicalldenedintermofintegraloftheoluonoverdomainor
oundarie.ItialopoiletoetupanojecvefunconaaProeataparcularlocaon
inthemodelpace.Anderivedquantaedontheoluoneld,uchathepaal
gradientoftheoluon,caneuedapartoftheojecvefuncon.

CloingRemark
Compungdeignenivieihelpfulfordeterminingwhichparameteraectour
ojecvefunconthemotandgiveuanideaaoutwhichparameterwemightwantto
focuonawetarttoconiderdeignchange.omeotherexamplethatuethi
funconalitincludeourtutorialmodelofanaxialmagnecearing
(hp://www.comol.com/model/axialmagnecearinguingpermanentmagnet14367)
andourenivitanaliofacommunicaonmatdetail
(hp://www.comol.com/model/enivitanaliofacommunicaonmatdetail4737).
Ofcoure,thimethodcaneuedinfarmorecaethanwecandecrieatonce.
Thitorconnuewhenwetarttouetheeenivietoimproveourojecve
funconwhereweopmizethedeign.ThicanedonewiththeOpmizaonModule
(hp://www.comol.com/opmizaonmodule),whichwewillcoverinanupcominglogpot.

Categorie
CoreFunconalit (/log/categor/all/corefunconalit/)

PotTag
TechnicalContent(/log/tag/technicalcontent/)

NewerPot
AvoidinghipHullCorroionwithICCPandimulaon
(hp://www.comol.com/log/avoidinghiphullcorroionwithiccpandimulaon/)

OlderPot
houldWeModelGrapheneaa2DheetorThin3DVolume?
(hp://www.comol.com/log/houldwemodelgrapheneaa2dheetorthin3d
volume/)

Comment
IvarKjelerg Augut6,20152:07am

HiWalter
Thankagainforaniceandclearexample,thelogpagearereallhelpfulandreplace
wellthelongmemiingTip&Trickoftheoldv3COMOLNew.
WhatIcouldnotetfromourhortdecriponiifouare,with,calculangthe
capacitanceincludingthefringeeldeectaroundthecapacitorplate.
Theperfectmatchougethowclearlthatenivitcalculaonworkperfectlwhen
weapplpuregeometricaldierenalequaon,aexpected.
ut,howwelldoetheimpliedgeometricalmodelmatchwhenoneincludetheelectric
eldcontriuonaroundthecapacitorplate?ThiiwhenFMi/houldecloertothe
realitthanimpliedgeometricalcalculaon.
incerel
Ivar
WalterFrei Augut6,20159:28am
DearIvar,
Thankouforourcomment.
Jutincaeitwanotcompletelclear:COMOLuetheniteelementmethodandthe
adjointmethodtocomputetheeenivie,otherearenopuregeometrical
dierenalequaoneinguedinthi,oranother,example.
Iimplchoeanexamplewhichcouldecomparedtoaneailderivaleanalc
oluon.Theniteelementaifunconuedhereareapplicalefor*an*geometr,
utofcoureIleaveittoourdearreadertoverifthi.
et,
MalkhazMeladze Augut6,20158:13pm
DearWalter,
IagreewithIvar,theComollogarealwahelpfulandexamplearealwachoen
thoughull.
ThiloginotanexceponIlearnedmuchfromit.Thankou.
Irepeatedtheexampleindierentvariaonandgettheamenumeradiplaedinthe
taleaoveonlifIpecifPrecriedMehDiplacement()adL/2anddD/2.Iam
wonderingwhatIamdoingwrong.AndapparentlIamdoingomethingwrong,ecaue
whatevervariaonIrun,ineachcaeIgettheamewarning:Newcontraintforcenode
detected:Theearenottored.
etregard,
Malkhaz
WalterFrei Augut7,20158:54am

DearMalkhaz,
Ye,therewilleawarningwhenourunthimodel,utoucanignoreit.Theenivit
reultthemelvearevalid.
Withrepecttoourreult:itoundlikeouareapplingthedeformaononothide
ofthedomain.Youcanaloetuponeidetohaveadiplacementofzero,andtheother
tohaveadiplacementofdLanddD.Thatwa,ouavoidthedividetwo.
et,
AlexeTrukhin eptemer12,20152:39am
DearWalter,
Thankouvermuchforthelog.
Couldoupleaeelaorateonwhaenivitanalicannotedonewitharaopc
module?
Kindregard,
WalterFrei eptemer22,201510:10am
DearAlexe,
TheRaOpcModuleuearatracingapproach(olvingODforthera)ratherthan
uingtheniteelementapproach(olvingPDfortheeld)andtheadjointenivit
calculaoninCOMOLonlworkwithprolemolvedviatheniteelementmethod.
(Ofcoure,inthefuturewemaintroduceetmoremethod.)
AlexeTrukhin eptemer24,201512:28pm
DearWalter,
IfIundetandoucorrectl,thereinoantoolforaenivitanaliintheRaOpc
moduleinthecurrentverionofCOMOL(verion5.1)?
WalterFrei eptemer24,20151:28pm
DearAlexe,
Itwouldemorefairtoathattheadjointenivitcalculaonwhichidecriedhere
ionlavailaleforniteelementaedprolem.TheRaOpcModuleiuiltupon
dierenttechnolog.Youcan,ofcoure,uetheParametricweepfunconalittooerve
thevariaonofanojecvefunconwithrepecttodeignparameterchange.
AlexeTrukhin eptemer26,20152:35am
DearWalter,
Thankouvermuchforclarifingtheituaon.Theenivitanaliimuchneededin
optomechanicaleldduetothefactthatthemotopcaldevice(aopcalightand
inocular)havetoeadjuted(adjungmirrororprim,forexample).Thatiwhiti

preferaletohaveaenivitanaliintheRaOpcmodule.
However,IhavenotfoundanretricononuingtheOpmizaonmoduleintheRa
Opcmoduleinthedocumentaon.ut,aumingthatothenivitandOpmizaon
modulearealliedtoomeextent,canwedeclarethattheOpmizaonmodulecannote
ulizedintheRaOpcmoduleeithertakingintoconideraonthattheOpmizaon
moduleuePD`forolvingopmizaonprolem?
WalterFrei eptemer28,20158:49am
DearAlexe,
YoucanuetheOpmizaonModulewiththeRaOpcModule,ouwouldimplhave
touethenumericalderivaveapproach,ratherthantheanalcderivaveapproach
developedhere.Ifoudohavedetailedqueonaoutthi,andareconideringthe
OpmizaonModuleforourmodelingneed,weencourageoutocontactourupport
ta,athiconveraonontheRaOpcModuleigengaiteondthecopeof
thilog.

Loginelowtoleaveacomment

LOGINTOYOURCOMOLACCACCOUNT
mail

aif.manji@koauniverit.org
Paword

Rememerme

Forgotpaword?()
Don'thaveanaccount?ignupnow()

LOGIN

NorthAmerica

(hp://plu.google.com/u/0/+comol/pot)

(//www.comol.com/log/)

(hp://www.faceook.com/mulphic)

Login(/acce/)
PrivacPolic(/compan/privac/) Trademark(/trademark/)
2016COMOLInc.Allrightreerved

(hp://www.linkedin.com/compan/comolinc)
(hp://twier.com/COMOL_Inc)

You might also like