数学物理学报, 2020, 40(3): 717-724 doi:

论文

确定热方程未知源问题的超阶正则化方法

赵振宇,1,2,3, 林日光2, 李志2, 梅端2

A Super Order Regularization Method for Determination of an Unknown Source in the Heat Equation

Zhao Zhenyu,1,2,3, Lin Riguang2, Li Zhi2, Mei Duan2

通讯作者: 赵振宇, E-mail: wozitianshanglai@163.com

收稿日期: 2019-02-18  

基金资助: 南方海洋科学与工程广东省实验室(湛江)资助项目.  ZJW-2019-04
广东海洋大学创新强校工程项目.  Q18306

Received: 2019-02-18  

Fund supported: the Fund of Southern Marine Science and Engineering Guangdong Laboratory (Zhanjiang).  ZJW-2019-04
the Project of Enhancing School with Innovation of Guangdong Ocean University.  Q18306

摘要

该文研究了热传导方程中未知源的确定问题.针对问题的不适定性,提出了一种结合超阶惩罚项的Tikhonov正则化方法.在由偏差原理选取正则化参数情况下,方法能够在不同光滑条件下获得最优收敛阶.计算过程不需要事先知道光滑度和精确解的先验界.数值试验表明,该方法是有效和稳定的.

关键词: 不适定问题 ; 超阶正则化 ; 热方程 ; 未知源 ; 偏差原理

Abstract

The problem for determining an unknown source in the heat equation is considered in this paper. We present a Tikhonov regularization method with a super order penalty term to deal with illposedness of the problem. The regularization parameter is chosen by a discrepancy principle and the order optimal error bounds can be obtained for various smooth conditions. The smoothness parameter and the a priori bound of exact solution are not needed in the numerical process. Numerical tests show that the proposed method is effective and stable.

Keywords: Ill-posed problem ; Super order regularization ; Heat equation ; Unknown source ; Discrepancy principle

PDF (345KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

赵振宇, 林日光, 李志, 梅端. 确定热方程未知源问题的超阶正则化方法. 数学物理学报[J], 2020, 40(3): 717-724 doi:

Zhao Zhenyu, Lin Riguang, Li Zhi, Mei Duan. A Super Order Regularization Method for Determination of an Unknown Source in the Heat Equation. Acta Mathematica Scientia[J], 2020, 40(3): 717-724 doi:

1 引言

本文考虑如下热方程未知源问题[6]

$ \begin{equation} \left\{ \begin{array}{llll} u_t = u_{xx}+f(x), &x\in {\Bbb R}, &0<t<1, \\ u(x, 0) = 0, &x\in {\Bbb R}, \\ u(x, 1) = g(x), & x\in {\Bbb R}, \end{array} \right. \end{equation} $

其中$u(\cdot, t)\in L^2({\Bbb R})$表示状态变量.我们的目标是由数据$u(x, 1)=g(x)$重构源项$f(x)$.实际过程中数据$g(x)$通常由物理装置读取获得.因此我们只能获得它的扰动数据$g^{\delta}(x)$.假设$g^{\delta}(x)\in L^2({\Bbb R})$且满足

$\begin{equation}\label{errorcondition} \|g-g^{\delta}\|\leq \delta, \end{equation}$

这里$\delta>0$表示误差水平, $\|\cdot\|$$L^2$ -范数.

反源问题在热传导、裂纹识别、电磁理论、污染源识别等许多物理和工程学科中有着广泛的应用.这些问题的主要困难是数值不稳定性.因此,在数值模拟过程中需要引入一些特殊的正则化手段,对这类问题研究已经有众多学者做了大量的工作.文献[2-3, 5, 7, 10, 15]中对解的唯一性和条件稳定性进行了研究.此类问题的数值重构方法可以参见文献[1, 6, 9, 11, 13-14, 16-20].

对本文问题而言,利用傅里叶变换,可以得到问题(1.1)的解如下[6]:

$\begin{equation}\label{fsol} f(x)={\cal F}^{-1}\left[\lambda(\xi)\hat{g}(\xi)\right]=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty} \frac{\xi^2e^{\xi^2}}{e^{\xi^2}-1}\hat{g}(\xi)e^{{\rm i}\xi x}{\rm d}x=:Tg, \end{equation}$

这里

$\begin{equation} \hat{g}(\xi)={\cal F}[g(x)]=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{+\infty}g(x)e^{-{\rm i}\xi x}{\rm d}x\end{equation} $

g的傅里叶变换,其中

$\lambda(\xi)=\frac{\xi^2e^{\xi^2}}{e^{\xi^2}-1}.$

显而易见, $\hat{g}$的衰减速率一定要超过$|\xi|^{-2}$.但是测量数据$g^{\delta}$一般不具备这样的衰减性质,因此我们无法通过传统方法获得问题解的合理近似.本文中我们将提出一种结合超阶罚项的吉洪诺夫正则化方法来处理不适定性.在此之前,我们首先对未知源设定先验性条件:

$\begin{equation}\label{prioriB} \|f\|_{p}\leq E, p\geq 0, \end{equation}$

此处$\|\cdot\|_p$表示索伯列夫空间$H^p({\Bbb R})$意义下的范数,可由下式定义

$\begin{equation}\label{boundfE} \|f\|_p:=\left(\int_{-\infty}^{+\infty}(1+\xi^2)^p\left|\hat{f}(\xi)\right|^2{\rm d}\xi\right)^{1/2}.\end{equation}$

文献[6]中给出了问题(1.1)的最优误差界,同时提出了一种结合先验参数准则的傅里叶正则化方法,在该方法中需要先验界$E$以及光滑度$p$的信息来确定正则化参数.文献[21]中提出了一种结合后验参数准则的修正吉洪诺夫正则化方法,但是该方法仍然需要参数$p$(通常未知)的信息.进一步文献[22]的作者通过希尔伯特尺度下的吉洪诺夫正则化方法来求解(1.1),在该文中,采用如下吉洪诺夫泛函的极小元作为$g$的近似:

$ \begin{equation} \Phi(\varphi)=\|\varphi-g^{\delta}\|^2+\alpha\|T\varphi\|^2, \end{equation}$

其中$\alpha>0$为正则化承诺书, $q$为正实数.该方法不需要预先知道$p$$E$的值,其理论收敛率为$O(\delta^{\frac{p}{p+2}}), p\leq 2q+2$.也就是说当$p\leq 2q+2$时方法时按阶最优的.对于较大的$p$值,由于浮点数的限制理论结果难以实现.

在文献[22]中, $\left\|T\varphi\right\|_q^2$被用作吉洪诺夫泛函中的罚项.从希尔伯特尺度正则化的观点来看,它是一种高阶正则化方法.本文将提出一种新的罚项(称作超阶正则化),理论结果表明,当采用偏差原理选取正则化参数时,对于任意的$p\in {\Bbb R}$方法均能获得最优收敛阶.我们已经在文献[4]中用类似的方式处理了数值微分问题.

本文安排如下:第2节给出了带超阶罚项的吉洪诺夫正则化方法的构造及一些辅助结果.正则化参数的选取和相应的收敛结果可以在第3节中找到.第4节通过数值实验验证了该方法的有效性.

2 求解问题(1.1)的结合超阶罚项的吉洪诺夫正则化方法

定义

$\begin{equation} \left({\cal R}\varphi\right)(x)={\cal F}^{-1}\left[e^{\xi}\hat{\varphi}(\xi)\right]=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty} e^{|\xi|}\hat{\varphi}(\xi)e^{{\rm i}\xi x}{\rm d}x, \end{equation}$

以及

$\begin{equation} \left({\cal P}_N\varphi\right)(x)=\frac{1}{\sqrt{2\pi}}\int_{-N}^{N}\hat{\varphi}(\xi)e^{{\rm i}\xi x}{\rm d}x. \end{equation}$

$\varphi^{\delta}=\varphi^{\alpha, \delta}$为如下吉洪诺夫泛函极小元

$\begin{equation}\label{TF} \Phi(\varphi)=\|\varphi-g^{\delta}\|^2+\alpha\left\|{\cal R}\varphi\right\|^2, \end{equation}$

其中$\alpha>0$为正则化参数. ${f}^{\alpha, \delta}=T{\varphi}^{\alpha, \delta}$将作为$f$的近似.

可以导出$\varphi^{\alpha, \delta}$为如下方程的解[8]

$\begin{equation}\label{TR} (I+\alpha{\cal F}^{-1}e^{2\xi}{\cal F})\varphi=g^{\delta}.\end{equation}$

可得

$\begin{equation} \hat{\varphi}^{\alpha, \delta}=\frac{1}{1+\alpha e^{2|\xi|}}\hat{g}^{\delta}.\end{equation}$

因此如果定义

$\begin{equation} \lambda(\xi)=\frac{\xi^2e^{\xi^2}}{e^{\xi^2}-1}, \end{equation}$

则正则化解${f}^{\alpha, \delta}$可以给出为

$\begin{equation}\label{sol1} {f}^{\alpha, \delta}=T{\varphi}^{\alpha, \delta}= {\cal F}^{-1}\left[\frac{\lambda(\xi)}{1+\alpha e^{2|\xi|}}\hat{g}^{\delta}\right].\end{equation}$

本文借鉴了文献[12]推导高阶方法的过程来建立超阶方法的框架,首先给出一些辅助性结果.

引理2.1[12]  假设

$ \begin{equation} k_\alpha(\lambda)=\frac{1}{\lambda+\alpha}, \end{equation}$

那么我们有

$\begin{equation} \sup\limits_{\lambda>0}\lambda^{1/2}|k_{\alpha}(\lambda)|\leq \frac{1}{2\sqrt{\alpha}}, \quad \sup\limits_{\lambda>0}\lambda|k_{\alpha}(\lambda)|\leq 1, \end{equation}$

以及

$\begin{equation} \sup\limits_{\lambda>0}\lambda^{1/2}|1-\lambda k_{\alpha}(\lambda)|\leq \frac{\sqrt{\alpha}}{2}, \quad \sup\limits_{\lambda>0}|1-\lambda k_{\alpha}(\lambda)|\leq 1.\end{equation}$

引理2.2[12]  对于任意$\xi\in {\Bbb R}$,我们有

$ \begin{equation}\label{xiineq} 0<\xi^2\leq \lambda(\xi)\leq 1+\xi^2\leq 2\lambda(\xi), \quad \xi\in {\Bbb R}. \end{equation}$

引理2.3  如果条件(1.6)成立,则

$ \begin{equation}\label{estpR} \|g-g_{N}\|\leq 2 {N}^{-p-2}E ~~~~~\hbox{及}~~~~~~~\|{\cal R} g_{N}\| \leq C_{N}E, \end{equation}$

其中

$ \begin{equation}\label{ctau} C_{N}=\max\left(1, \frac{2e^{N}}{N^{p+2}}\right). \end{equation}$

  由Parseval公式

$ \begin{eqnarray} \|g-g_{\tau}\|^2&=&\|\hat{g}-\hat{g}_{\tau}\|^2=\displaystyle{\int_{|\xi|\geq N}\hat{g}^2(\xi){\rm d}\xi} \\ &\leq& \displaystyle{\int_{|\xi|\geq N}\frac{4\lambda^2(\xi)}{(1+\xi^2)^2}\hat{g}^2(\xi){\rm d}\xi} \\ &=&\displaystyle{\int_{|\xi|\geq N}\frac{4(1+\xi^2)^p\lambda^2(\xi)}{(1+\xi^2)^{p+2}}\hat{g}^2(\xi){\rm d}\xi} \\ &\leq &\displaystyle{\frac{4}{\tau^{2p+4}}\int_{|\xi|\geq N}(1+\xi^2)^p\hat{f}^2(\xi){\rm d}\xi} \\& \leq& \displaystyle{ \frac{4}{N^{2p+4}}\|f\|^2_p}, \end{eqnarray}$

以及

$ \begin{eqnarray} \|{\cal R} g_{N}\|^2&=&\|e^{|\xi|}\hat{g}_{N}\|^2=\displaystyle{\int_{|\xi|\leq N}e^{2|\xi|}\hat{g}^2(\xi){\rm d}\xi} \\ &\leq &\displaystyle{\int_{|\xi|\geq N}\frac{4e^{2|\xi|}\lambda^2(\xi)}{\xi^4}\hat{g}^2(\xi){\rm d}\xi} \\ &\leq&\displaystyle{\int_{|\xi|\geq N}\frac{4e^{2|\xi|}}{\xi^{2p+4}}(1+\xi^2)^p\hat{f}^2(\xi){\rm d}\xi} \\ &\leq &\max\left(1, \frac{4e^{2N}}{N^{2p+4}}\right)\|f\|_p^2. \end{eqnarray}$

引理2.3证毕.

引理2.4[21]  如果$\|T\varphi\|_p$有界,则有

$\begin{equation} \|T\varphi\|\leq \|\varphi\|^{\frac{p}{p+2}}\|T\varphi\|_p^{\frac{2}{p+2}}.\end{equation}$

引理2.5  假设函数族$\varphi^{\delta}$满足

$\begin{equation}\label{stablecondition} \|\varphi^{\delta}\|\leq k_1\delta~~~~~\hbox{和}~~~~~~\|{\cal R}\varphi^{\delta}\|\leq k_2e^{k_3\delta^{-\frac{1}{p+2}}}\delta , ~~~~~~~\delta\rightarrow 0, \end{equation}$

其中$k_1, k_2, k_3$为固定实数.则存在实数$M$满足

$ \begin{equation} \|T\varphi^{\delta}\|_p\leq M, ~~~~~~~~~~~~~~\delta\rightarrow0. \end{equation}$

  可以证明存在一个常数$\delta_0$满足

$\begin{equation}\label{estdandiao} e^{2k_3\delta^{-\frac{1}{p+2}}}>\frac{(2k_3)^{p+2}}{\delta}, \quad \forall \delta<\delta_0.\end{equation}$

不失一般性,我们仅证明$\delta<\delta_0$情况.令

$\begin{equation} N_0=k_3\delta^{-\frac{1}{p+2}}, \end{equation}$

则由三角不等式

$\begin{equation} \|T\varphi^{\delta}\|_p \leq \|T(I-{\cal P}_{N_0})\varphi\|_p+\|T{\cal P}_{N_0}\varphi\|_p=:I_1+I_2. \end{equation}$

现在我们分别来估计$I_1$$I_2$.对于第一项$I_1$,由(2.11), (2.12), (2.19)及(2.17)式可得

$ \begin{eqnarray} I^2_1&=&\displaystyle{\int_{|\xi|> N_0}(1+\xi^2)^p\lambda^2(\xi)\left[\hat{\varphi}^{\delta}(\xi)\right]^2{\rm d}\xi} \\ &\leq&{\int_{\|\xi\|> N_0}\frac{(1+\xi^2)^{p+2}}{e^{2|\xi|}}\left[e^{\xi}\hat{\varphi}^{\delta}(\xi)\right]^2{\rm d}\xi} \\& \leq& \frac{(2\tau)^{2p+4}}{e^{2N_0}}\|{\cal R}\varphi\|^2=2^{2p+4}k_2^2k_3^{2p+4}. \end{eqnarray}$

对于第二项$I_2$,结合(2.11)及(2.17)式有

$ \begin{eqnarray} I^2_2&=&\displaystyle{\int_{|\xi|\leq N_0}(1+\xi^2)^p\lambda^2(\xi)\left[\hat{\varphi}^{\delta}(\xi)\right]^2{\rm d}\xi} \\ &\leq&\int_{|\xi|\leq N_0}(1+\xi^2)^{p+2}\left[\hat{\varphi}^{\delta}(\xi)\right]^2{\rm d}\xi \\ &\leq &4N_0^{2p+4}\int_{|\xi|\leq N_0}\left[\hat{\varphi}^{\delta}(\xi)\right]^2{\rm d}\xi \\ &\leq &4N_0^{2p+4}\|\varphi\|^2 =4k_1^2k_3^{2p+4}. \end{eqnarray}$

引理2.5证毕.

定理2.1  假设条件(1.2), (1.6)成立, ${f}^{\alpha, \delta}$由(2.7)式定义,如果我们选取参数$\alpha$满足

${\|\varphi^{\alpha, \delta}-g^{\delta}\|}=C\delta, $

其中$C>1$,则

$ \begin{equation} \|{f}^{\alpha, \delta}-f\|=2E^{\frac{2}{p+2}}\delta^{\frac{p}{p+2}}.\end{equation}$

  令${\cal T}={\cal R}^{-1}$,那么

$\begin{equation}\label{regularizationsolution} {\varphi}^{\alpha, \delta}={\cal T}k_{\alpha}({\cal T}^2){\cal T}g^{\delta}. \end{equation}$

定义$g_{\alpha, N}={\cal T}k_{\alpha}({\cal T}^2){\cal T}{g}_{N}$,则有

$ \begin{equation} \begin{array}{lllllll} ({\varphi}^{\alpha, \delta}-{g}_{\alpha, N})={\cal T}k_{\alpha}({\cal T}^2){\cal T}(g^{\delta}-{g}_N), \\ ({g}_{N}-{g}_{\alpha, N})={\cal T}[I-k_{\alpha}({\cal T}^2){\cal T}^2]{\cal R}{g}_{N}, \\ {\cal R}({\varphi}^{\alpha, \delta}-{g}_{\alpha, N})=k_{\alpha}({\cal T}^2){\cal T}(g^{\delta}-{g}_{N}), \\ {\cal R}({g}_{N}-{g}_{\alpha, N})=[I-k_{\alpha}({\cal T}^2){\cal T}^2]{\cal R}{g}_{N}. \end{array} \end{equation}$

进而

$ \begin{eqnarray}\label{boundf} \|({\varphi}^{\alpha, \delta}-{g}_{N})\|&\leq& \|{\varphi}^{\alpha, \delta}-g^{\delta}\|+ \|g^{\delta}-g\|+ \|g-{g}_{N}\| \\ &\leq& (C+1)\delta+\|g-g_{N}\|\leq(C+1)\delta+2N^{-p-2}E, \end{eqnarray}$

以及

$ \begin{eqnarray}\label{boundR} \|{\cal R}({\varphi}^{\alpha, \delta}-{g}_{N})\|_{l^2}&\leq& \|{\cal R}({\varphi}^{\alpha, \delta}-{g}_{\alpha, N})\|_{l^2}+ \|{\cal R}({g}_{\alpha, N}-{g}_{N})\|_{l^2} \\ &\leq &\frac{1}{2\sqrt{\alpha}}\|g^{\delta}-{g}_{N}\|+\|{\cal R}{g}_{N}\|_{l^2} \\ &\leq& \frac{1}{2\sqrt{\alpha}}(\delta+2N^{-p-2}E)+C_N E. \end{eqnarray}$

${\cal S}_{\alpha}=I-k_{\alpha}({\cal T}^2){\cal T}^2$,由表达式$g^{\delta}-{\varphi}^{\alpha, \delta}={\cal S}_{\alpha}g^{\delta}$可得

$ \begin{eqnarray} C\delta &=&{\|{\varphi}^{\alpha, \delta}-g^{\delta}\|} \\ &\leq &{\|{\cal S}_{\alpha}(g^{\delta}-g)\|}+{\|{\cal S}_{\alpha}(g-g_N)\|}+{\|{\cal S}_{\alpha}g_{N}\|} \\ &\leq &\delta+{\|{\cal S}_{\alpha}(g-g_{N})\|}+{\|{\cal S}_{\alpha}g_{N}\|} \\ &\leq &\delta+{\|(g-g_{N})\|}+{\|{\cal S}_{\alpha}{\cal T}\|}\cdot{\|{\cal R}{g}_{N}\|} \\ &\leq& \delta+2N^{-p-2}E+\frac{\sqrt{\alpha}}{2}C_{N}E. \end{eqnarray}$

如果选取$N$满足

$ \begin{equation}\label{choicep} N^{-p-2}E=\frac{C-1}{4}\delta, \end{equation}$

则有

$ \begin{equation}\label{eq240} \frac{(C-1)\delta}{\sqrt{\alpha}}\leq C_{N}E. \end{equation}$

由(2.13), (2.28), (2.29), (2.31)及(2.32)式,有

$ \begin{equation} \|{\varphi}^{\alpha, \delta}-{g}_{N})\|\leq \frac{3C+1}{2}\delta, \end{equation}$

$ \begin{equation} \|{\cal R}({\varphi}^{\alpha, \delta}-{g}_{N})\|_{l^2}\leq \frac{C-1}{2}e^{\left(\frac{4E}{C-1}\right)^{\frac{1}{p+2}}\delta^{-\frac{1}{p+2}}}\delta. \end{equation}$

则存在常数$M$满足

$ \begin{equation} \|T({\varphi}^{\alpha, \delta}-{g}_{N})\|_p\leq M. \end{equation}$

所以有

$ \begin{eqnarray}\label{final1} \|T({\varphi}^{\alpha, \delta}-g)\|_p &\leq& \|T({\varphi}^{\alpha, \delta}-{g}_{N})\|_p+\|T({g}-{g}_{N})]\|_p \\ &\leq& \|T({\varphi}^{\alpha, \delta}-{g}_{N})\|_p+\|f\|_p \leq M+E. \end{eqnarray}$

进一步由(1.2)式, (2.24)式及三角不等式得

$ \begin{equation}\label{final2} \|g^{\alpha, \delta}-g\|\leq\|{\cal F}{\bf v}_{\alpha}^{\delta}-g^{\delta}\|+\|{g}^{\delta}-g\|\leq (C+1)\delta.\end{equation}$

由(2.36)式, (2.37)式及引理2.4即可得到结论.

3 数值例子

本文方法可以由快速傅里叶变换实现,考虑如下例子.

[6]  容易验证函数对

$\begin{equation} u(x, t)=\frac{x}{(t+1)^{\frac{3}{2}}}\exp\left(-\frac{x^2}{4(t+1)}\right)-x\exp\left(-\frac{x^2}{4}\right), \end{equation}$

$\begin{equation} f(x)=\left(\frac{x^3}{4}-\frac{3x}{2}\right)\exp\left(-\frac{x^2}{4}\right)\end{equation}$

为问题(1.1)当取

$\begin{equation} g(x)=\frac{x}{2^{\frac{3}{2}}}\exp\left(-\frac{x^2}{8}\right)-x\exp\left(-\frac{x^2}{4}\right)\end{equation}$

时的解.因为当$|x|>6$$f(x)$接近于0,我们截取区间$-10\leq x\leq 10$进行数值实现.设$x_j=-10+jh, j=0, 1, \cdots, N-1, N=256$.扰动数据给定为

$\begin{equation} g^{\delta}(x_j)=g(x_j)+\epsilon_j, \end{equation}$

其中$\{\epsilon_j\}_{j=0}^{N}$由Matlab函数$\hbox{randn}(N, 1)\times\delta_1$生成.我们将把本文方法(M1)与文献[22]方法(M2)进行对比. 表 1给出了相应的相对误差,可以看出随着$\delta_1$$0.1$下降到$0.0001$,误差$\|f^{\bar{\alpha}, \delta}-f\|$变小且方法M1结果优于方法M2.

表 1   数值结果展示

$\delta_1$M2M1
$q=2$$q=4$$q=8$
1e-19.18e-28.68e-28.90e-29.68e-2
1e-21.81e-21.54e-21.47e-21.41e-2
1e-32.91e-32.24e-31.98e-31.58e-3
1e-45.99e-42.81e-42.25e-41.75e-4

新窗口打开| 下载CSV


4 结论

本文提出了一种识别热方程未知源问题的吉洪诺夫方法,引入了新的罚项改进了以往方法的一些缺点.理论显示方法对于任意光滑度均是按阶最优的.数值实验进一步验证了方法的有效性.

参考文献

Burykin A A , Denisov A M .

Determination of the unknown sources in the heat-conduction equation

Comput Math Model, 1997, 8 (4): 309- 313

URL     [本文引用: 1]

Cannon J R , DuChateau P .

Structural identification of an unknown source term in a heat equation

Inverse Probl, 1998, 14 (3): 535- 551

URL     [本文引用: 1]

Cannon J R , Perez-Esteva S .

Uniqueness and stability of 3d heat sources

Inverse Probl, 1991, 7, 57- 62

DOI:10.1088/0266-5611/7/1/006      [本文引用: 1]

Chen B Q , Zhao Z Y , Li Z , Meng Z H .

Numerical differentiation by a fourier extension method with super-order regularization

Appl Math Comput, 2018, 334, 1- 10

DOI:10.1016/j.cam.2017.11.024      [本文引用: 1]

Choulli M , Yamamoto M .

Conditional stability in determining a heat source

J Inverse Ill-Posed Probl, 2004, 12 (3): 233- 243

URL     [本文引用: 1]

Dou F F , Fu C L , Yang F L .

Optimal error bound and Fourier regularization for identifying an unknown source in the heat equation

J Comput Appl Math, 2009, 230 (2): 728- 737

URL     [本文引用: 5]

Badia A E , Duong T H .

On an inverse source problem for the heat equation. Application to a pollution detection problem

J Inverse Ill-Posed Probl,, 2002, 10 (6): 858- 600

URL     [本文引用: 1]

Engl H W , Hanke M , Neubauer A . Regularization of Inverse Problems. Netherlands: Springer, 1996

[本文引用: 1]

Farcas A , Lesnic D .

The boundary-element method for the determination of a heat source dependent on one variable

J Engrg Math, 2006, 54 (4): 375- 388

DOI:10.1007/s10665-005-9023-0      [本文引用: 1]

Li G S .

Data compatibility and conditional stability for an inverse source problem in the heat equation

Appl Math Comput, 2006, 173 (1): 566- 581

URL     [本文引用: 1]

Ling L , Yamamoto M M , Hon Y C , Takeuchi T .

Identification of source locations in two-dimensional heat equations

Inverse Probl, 2006, 22 (4): 1289- 1305

URL     [本文引用: 1]

Nair M T , Pereverzev S V , Tautenhahn U .

Regularization in hilbert scales under general smoothing conditions

Inverse Problems, 2005, 21 (21): 1851- 1862

URL     [本文引用: 3]

Park H M , Chung J S .

A sequential method of solving inverse natural convection problems

Inverse Probl, 2002, 18 (3): 529- 546

URL     [本文引用: 1]

Ryabenḱii V S , Tsynkov S V , Utyuzhnikov S V .

Inverse source problem and active shielding for composite domains

Appl Math Lett, 2007, 20 (5): 511- 515

DOI:10.1016/j.aml.2006.05.019      [本文引用: 1]

Yamamoto M .

Conditional stability in determination of force terms of heat equations in a rectangle

Math Comput Modelling, 1993, 18 (1): 79- 88

URL     [本文引用: 1]

Yan L , Fu C L , Yang F L .

The method of fundamental solutions for the inverse heat source problem

Eng Anal Bound Elem, 2008, 32 (3): 216- 222

DOI:10.1016/j.enganabound.2007.08.002      [本文引用: 1]

Yan L , Yang F L , Fu C L .

A meshless method for solving an inverse spacewise-dependent heat source problem

J Comput Physics, 2009, 228 (1): 123- 136

URL    

Yang F .

The truncation method for identifying an unknown source in the poisson equation

Appl Math Comput, 2011, 22, 9334- 9339

URL    

Yang F , Fu C L .

The modified regularization method for identifying the unknown source on poisson equation

Appl Math Modelling, 2012, 2, 756- 763

URL    

Yi Y , Murio D A .

Source term identification in 1-D IHCP

Computers Math Appl, 2004, 47 (12): 1921- 1933

DOI:10.1016/j.camwa.2002.11.025      [本文引用: 1]

Zhao Z Y , You L .

A modified Tikhonov regularization method for identifying an unknown source in the heat equation

Acta Math Sci, 2014, 34A, 186- 192

URL     [本文引用: 2]

Meng Z H , You L , Zhao Z Y , Xie O .

Determination of an unknown source in the heat equation by the method of Tikhonov regularization in hilbert scales

J Appl Math Physics, 2014, 2 (2): 42095

URL     [本文引用: 3]

/