简介：BasedonArnoldi’smethod,aversionofgeneralizedArnoldialgorithmhasbeendevel-opedforthereductionofgyroscopiceigenvalueproblems.Byutilizingtheskewsymmetryofsystemmatrix,averysimplerecurrencescheme,namedgyroscopicArnoldireductionalgorithmhasbeenob-tained,whichisevensimplerthantheLanczosalgorithmforsymmetriceigenvalueproblems.Thecomplexnumbercomputationiscompletelyavoided.Arestarttechniqueisusedtoenablethereductionalgorithmtohaveiterativecharacteristics.Ithasbeenfoundthattherestarttechniqueisnotonlyef-fectivefortheconvergenceofmultipleeigenvaluesbutitalsofurnishesthereductionalgorithmwithatechniquetocheckandcomputemissedeigenvalues.Bycombiningitwiththerestarttechnique,theal-gorithmismadepracticalforlarge-scalegyroscopiceigenvalueproblems.Numericalexamplesaregiv-entodemonstratetheeffectivenessofthemethodproposed.

简介：Asix-variablegeometricalnonlinearsheardeformationlaminatedtheoryispresentedbywhichnormalstressandstraindistributioncanbecalculated.Byconsideringsomeaffectivefactorsthatwereneglectedunderthefinitedeformationcondition,animprovedyonKarmangeometricalnon-lineardeformation-strainrelationisusedforlargedeformationanalysis.Afteranalyzingthebendingproblemoflaminatedplates,andcomparingitwith3-DelasticitysolutionsandJ.N.Reddyfive-vari-ablesimplehigher-ordersheardeformationlaminatedtheory,wecanconcludethatasatisfactorycal-culationprecisionhasbeenachieved,whichshowsthatitisespeciallysuitableforthecalculationintheconditionoflargedeformationandthelaminatedthickplateanalysis.

简介：Inthispaper,anewzigzagmethodforplatestructuresandageneticalgorithm(GA)ofdynamicsourceseedspacesaredevelopedandacombinationofthemisusedtodealwithlargescalebuilt-upstructuraloptimization.ThenewGAcombinedwiththezigzagmethodcanworkefficientlywhencopingwithlargescalestructuraloptimizationincludeddisplacementandstressconstraints.ExamplesshowthatthisGAisrobustandcanbeusedformanycomplexstructuraloptimizationproblems.

简介：Thecomputationalefficiencyofnumericalsolutionoflinearalgebraicequationsinfiniteelementscanbeimprovedintowwqys.Oneistodecreasethefill-innumbers,whicharenewnon-ze-ronumbersinthematrixofglobalstiffnessgeneratedduringtheprocessofelimination.Theotheristoreducethecomputationaloperationofmultiplyingarealnumberbyzero.Basedonthefactthattheorderofeliminationcandeterminehowmanyfill-innumbersshouldbegenerated,wepresentanewmethodforoptimizationofnumberingnodes.Thismethodisquitedifferentfrombandwidthoptimization.Fill-innumberscanbedecreasedinalargescalebytheuseofthismethod.Thebi-factorizationmethodisadotedtoavoidmultiplyingrealnumbersbyzero.Forlargescalefiniteelementanalysis,themethodpresentedinthispaperismoreefficientthanthetraditionalLDLTmethod.

简介：Thesingularboundarymethod(SBM)isarecentmeshlessboundarycollocationmethodthatremediestheperplexingdrawbackoffictitiousboundaryinthemethodoffundamentalsolutions(MFS).Thebasicideaistousetheoriginintensityfactortoeliminatesingularityofthefundamentalsolutionatsource.Themethodhassofarbeenappliedsuccessfullytothepotentialandelasticityproblems.However,theSBMsolutionforlarge-scaleproblemshasbeenhinderedbytheoperationcountofO(N~3)withdirectsolversorO(N~2)withiterativesolvers,aswellasthememoryrequirementofO(N~2).Inthisstudy,thefirstattemptwasmadetocombinethefastmultipolemethod(FMM)andtheSBMtosignificantlyreduceCPUtimeandmemoryrequirementbyonedegreeofmagnitude,namely,O(N).Basedonthecomplexvariablerepresentationoffundamentalsolutions,theFMM-SBMformulationsforbothdisplacementandtractionwerepresented.Numericalexampleswithuptohundredsofthousandsofunknownshavesuccessfullybeentestedonadesktopcomputer.TheseresultsclearlyillustratedthattheproposedFMM-SBMwasveryefficientandpromisinginsolvinglarge-scaleplaneelasticityproblems.

简介：Thispaperstudiesthedynamicbehavioroflargedeformationofsphericalshellsimpactedbyaflat-nosedmissile.Byusingisometrictransformations,thedeformationmodesaregiven.OnthebasisofPerzyna-Symondsviscoplasticconstitutiveequations,themotionequationsoftheshellsareobtainedbyrigid-viscoplasticvariationalprinciple.Acomparisonmadebetweenthenumericalresultsandexperimentalonesindicatesthatthetwogroupsofresultsareinconformitywitheachother.