|
BoundaryElementMethod(BEM)computermodelstypicallyinvolveuseofnodalpointsthatarethe locationsofsingularpotentialfunctionssuchasthelogarithmorreciprocaloftheEuclideandistance function.Thesesingularfunctionsaretypicallyassociatedwiththenodesthemselvesasfaras identification.TheComplexVariableBoundaryElementMethod(CVBEM)isanotherapplicationof similartypesofsingularpotentialfunctionsandincludesotherfunctionsthatarenotsingularbutare fundamentalsolutionsofthegoverningpartialdifferentialequation(PDE).Thesevarioussingular potentialfunctionsformabasiswhosespanoflinearcombinations(eitherrealorcomplexspace,as appropriate)isavectorspace.Aspartoftheapproximationapproach,onedeterminesthatelementin the vectorspacethatisclosest(usuallyinaleastsquaresresidualmeasure)totheexactsolutionofthe PDE andrelatedboundaryconditions.RecentresearchonthetypesofbasisfunctionsusedinaBEMor CVBEMapproximationhasshownthatconsiderableimprovementincomputationalaccuracyand efficiencycanbeachievedbyoptimizingthelocationofthesingularbasisfunctionswithrespectto possiblelocationsontheproblemboundaryandalsolocationsexterioroftheproblemboundary(in general,exterioroftheproblemdomain).Todevelopsuchoptimumlocationsforthemodelingnodes (and associatedsingularbasisfunctions),theapproachpresentedinthispaperistodevelopaRealTime BoundaryElementNodeLocationmodulethatenablestheprogramusertoclickanddragnodes(oneat a time)throughouttheexterioroftheproblemdomain(thatis,nodesareallowedtobepositionedon or arbitrarilyclosetotheproblemboundary,andalsotobepositionedexterioroftheproblemdomain union boundary).TheprovidedmoduleinterfaceswiththeCVBEMprogram,builtwithincomputer programMathematica,sothatvarioustypesofinformationflowstothedisplaymoduleasthenodeis moved,inrealtime.Theinformationdisplayedincludesagraphicoftheproblemboundaryand domain,theexteriorofthedomainunionboundary,evaluationpointsusedtorepresentproblem boundaryconditions,nodallocations,modelingerrorin L2 and also LN norms,andaplotofproblem boundaryconditionsversusmodelingestimatesontheproblemboundarytoenableavisualizationof closenessoffitofthemodeltotheproblemboundaryconditions.Asthetargetnodeismovedonthe screen,thesevariousinformationformschangeandaredisplayedtotheprogramuser,enablingthe usertoquicklynavigatethetargetnodetowardsapreferredlocation.Onceanodeisestablishedat someoptimizedlocation,anothernodecanthenbeclickeduponanddraggedtonewlocations,while reducingmodelingerrorintheprocess.
|
|