简介:Anadaptivemulti-scaleconjugategradientmethodfordistributedparameterestimations(orinverseproblems)ofwaveequationispresented.Theidentificationofthecoefficientsofwaveequationsintwodimensionsisconsidered.First,theconjugategradientmethodforoptimizationisadoptedtosolvetheinverseproblems.Second,theideaofmulti-scaleinversionandthenecessaryconditionsthattheoptimalsolutionshouldbethefixedpointofmulti-scaleinversionmethodisconsidered.Anadaptivemulti-scaleinversionmethodfortheinverseproblemisdevelopedinconjunctionwiththeconjugategradientmethod.Finally,somenumericalresultsareshowntoindicatetherobustnessandeffectivenessofourmethod.
简介:这份报纸是一个continue工作[4,5]。在以前的二份报纸,我们与看情况时间的抑制术语学习了多维的可压缩的Euler方程的Cauchy问题--(1+t)u,在此0并且>0是常数。我们显示出那,为所有0并且>0,Cauchy问题的光滑的答案全球性存在或在有限时间骤起。在现在的纸,而不是Cauchy问题,我们在一半考虑起始边界的价值问题空间有空间尺寸d=的d+2,3。在方程和液体涡度的特殊结构的帮助下,我们克服困难从边界效果产生。我们证明那在那里存在0的一个全球光滑的解决方案<1当起始的数据接近它的平衡状态时。另外,液体涡度的指数的腐烂将也被建立。
简介:Thenumericalsolutionoflargescalemulti-dimensionalconvectiondiffusionequationsoftenrequiresefficientparallelalgorithms.Inthiswork,weconsidertheextensionofarecentlyproposednon-overlappingdomaindecompositionmethodfortwodimensionaltimedependentconvectiondiffusionequationswithvariablecoefficients.Bycombiningpredictor-correctortechnique,modifiedupwinddifferenceswithexplicitimplicitcoupling,themethodunderconsiderationprovidesintrinsicparallelismwhilemaintaininggoodstabilityandaccuracy.Moreover,formulti-dimensionalproblems,themethodcanbereadilyimplementedonamulti-processorsystemanddoesnothavethelimitationonthechoiceofsubdomainsrequiredbysomeothersimilarpredictor-correctororstabilizedschemes.Thesepropertiesofthemethodaredemonstratedinthisworkthroughbothrigorousmathematicalanalysisandnumericalexperiments.