资源描述：
《基于神经网络解决有限元网格剖分》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、1482IEEETRANSACTIONSONNEURALNETWORKS,VOL.13,NO.6,NOVEMBER2002AFinite-ElementMeshGeneratorBasedonGrowingNeuralNetworksDimitrisG.TriantafyllidisandDimitrisP.Labridis,SeniorMember,IEEEAbstract—Ameshgeneratorfortheproductionofhigh-qualitydefinitions.Startingfromaninitialcoarsemesh,theproposedfinite-
2、elementmeshesisbeingproposed.Themeshgeneratorusesmeshgeneratorwillprovideaqualitymesh,whichwillserveasanartificialneuralnetwork,whichgrowsduringthetrainingagoodstartingpointfortheadaptiverefinementtofollow.processinordertoadaptitselftoaprespecifiedprobabilitydistri-Oneofthemostcommonmethodsofmes
3、hgenerationistobution.TheinitialmeshisaconstrainedDelaunaytriangulation(CDT)ofthedomaintobetriangulated.Twonewalgorithmstostartfromaninfinitetriangulargrid,whichisthensuperimposedacceleratethelocationofthebestmatchingunitareintroduced.ontheobjecttobemeshed.TheelementsthatfalloutsidetheThemeshgen
4、eratorhasbeenfoundabletoproducemeshesofhighobjectareeitherremovedortrimmedtofitthegeometryofthequalityinanumberofclassiccasesexaminedandishighlysuitedobject.Thismethodwillproducevery-high-qualityelementsinforproblemswherethemeshdensityvectorcanbecalculatedintheinteriorofthemeshingarea,butisnotsu
5、itableincaseswhereadvance.smallobjectsarepresentinalargesolvingregion.Suchprob-IndexTerms—Automaticmeshgeneration,bestmatchingunitlemsappearinoverheadpowertransmissionline(OTL)prob-location,finite-elementmethod(FEM),let-it-grow(LIG)neuralnetworks,meshdensityprediction.lems,wherethesizeofthecondu
6、ctorsisminimalcomparedtothesolvingregion.Theinitialgridwouldhavetobemadeofel-ementsofverysmallsizeinordertomatchthegeometryoftheI.INTRODUCTIONconductors.ThiswouldproduceanunnecessarylargenumberofHEFINITE-elementmethod(FEM)isoneofthemostelementsintherestofthesolvingarea.AhybridmethodcouldTwidelyu
7、sednumericalmethodsinengineering,especiallyprobablybeusedinwhichtheareatobemeshedwouldbesubdi-duetoitsabilitytocopewithproblemsofhighgeometricalvidedinsmallerregions,butthemethodwouldbecasespecificcomplexity,whereananalytica
