同时反演热传导问题初值和源项的两种正则化方法.pdf

摘要

摘要

本文讨论了热传导问题中通过=和=(0)两个时刻的给定

1212

数据来同时反演初始条件和未知源项.在该问题中对源项的反演是中度不适定的,

对初值的反演是严重不适定的,因为在频域空间中,高频元素的微小扰动将会引起

解的巨大误差.本文先分别用Fourier正则化方法和简化的Tikhonov正则化得到了

正则化近似解,又分别给出了正则化参数选取规则下Fourier正则化近似解与精确

解之间的对数型收敛估计,及简化的Tikhonov正则化近似解和精确解的¨型

估计.最后通过数值实验表明了这两种正则化方法对该问题是有效的.

关关关键键键词词词:不适定问题;热传导方程;正则化方法;正则化参数选取规则;误差估

计;数值实验.

I

摘要

Abstract

Inthispaper,wediscussedtheheatconductionproblemby=1,=2

(0)twotimegiven.Thedataareusedtosimultaneouslyinverttheinitial

12

conditionsandtheunknownsourceterm,inwhichtheinversionofthesourcetermis

moderatelyill-posed,Theinversionoftheinitialvalueisseriouslyill-posed,because

inthefrequencydomainspace,thesmallperturbationofthehighfrequencyelement

willcause.Inthispaper,wefirstuseFourierregularizationmethodandsimplified

Tikhonovregularizationtoobtainthelargeerrorofthesolution.Theapproximate

solutionofregularizationisgiven,andtheapproximatesolutionandaccuracyof

Fourierregularizationundertheregularizationparameterselectionrulearegiven

respectively.Thelogarithmicconvergenceestimatesbetweenthesolutions,andthe

¨typeofthesimplifiedTikhonovregularizationapproximatesolutionandthe

exactsolution.Finally,numericalexperimentsshowthatthesetworegularization

methodsareeffectiveforthisproblem.

Keywords:Ill-posedproblem;heatconductionequation;Regularizationmethod;

Selectionofregularizedparameter;Errorestimation;Numericalexperiment