数学物理学报, 2019, 39(4): 839-850 doi:

论文

时间分数阶扩散方程线性三角形元的高精度分析

史艳华,, 张亚东, 王芬玲,, 赵艳敏, 王萍莉

High Accuracy Analysis of Linear Triangular Element for Time Fractional Diffusion Equations

Shi Yanhua,, Zhang Yadong, Wang Fenling,, Zhao Yanmin, Wang Pingli

通讯作者: 王芬玲, E-mail: mathwfl@163.com

收稿日期: 2018-02-28  

基金资助: 河南省高等学校重点科研项目.  17A110011

Received: 2018-02-28  

Fund supported: the Key Scientific Research Projects in Universities of Henan Province.  17A110011

作者简介 About authors

史艳华,E-mail:syhsdq@163.com , E-mail:syhsdq@163.com

摘要

该文基于线性三角形元和改进的$L1$格式,对具有$\alpha$阶Caputo导数的时间分数阶扩散方程建立了一个全离散逼近格式.首先,证明了该格式的无条件稳定性.其次,利用该单元及Ritz投影算子的性质,导出了关于投影算子具有$O(h^2+\tau^{2-\alpha})$阶的超逼近性质.再结合插值算子和投影算子的关系,进一步导出了关于插值算子具有$O(h^2+\tau^{2-\alpha})$阶的超逼近性质.然后,借助插值后处理技术得到了整体超收敛估计.最后,利用数值算例验证了理论分析的正确性.

关键词: 时间分数阶扩散方程 ; 线性三角形元 ; 全离散格式 ; 无条件稳定 ; 超逼近和超收敛

Abstract

In this paper, based on linear triangular element and improved $L1$ approximation, a fully-discrete scheme is proposed for time fractional diffusion equations with $\alpha$ order Caputo fractional derivative. Firstly, the unconditional stability is proved. Secondly, by employing the properties of the element and Ritz projection operator, superclose analysis for the projection operator is deduced with order $O(h^2+\tau^{2-\alpha})$. Further more, combining with relationship between the interpolation operator and Ritz projection, superclose analysis for the interpolation operator is also investigated with order $O(h^2+\tau^{2-\alpha})$. And then, the superconvergence result is obtained through the interpolated postprocessing technique. Finally, numerical results are provided to show the validity of our theoretical analysis.

Keywords: Time fractional diffusion equations ; Linear triangular element ; Fully-discrete scheme ; Unconditional stability ; Superclose and superconvergence

PDF (414KB) 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

史艳华, 张亚东, 王芬玲, 赵艳敏, 王萍莉. 时间分数阶扩散方程线性三角形元的高精度分析. 数学物理学报[J], 2019, 39(4): 839-850 doi:

Shi Yanhua, Zhang Yadong, Wang Fenling, Zhao Yanmin, Wang Pingli. High Accuracy Analysis of Linear Triangular Element for Time Fractional Diffusion Equations. Acta Mathematica Scientia[J], 2019, 39(4): 839-850 doi:

1 引言

含有分数阶导数的偏微分方程称为分数阶偏微分方程,它分为时间分数阶微分方程、空间分数阶微分方程和时空分数阶微分方程.对于时间分数阶微分方程,根据时间分数阶$\alpha$的大小,又可以分为扩散分数阶微分方程($0<\alpha<1$)和对流分数阶微分方程($1<\alpha<2$).分数阶导数具有非局部性,较之整数阶微分方程,它能更为精确地刻画出在时间和空间上的某种非局部关系,很多数学问题用分数阶微分方程描述更为合适.因此它已被广泛应用于许多学科和领域,如粘弹性力学、流体和复杂网络中的幂律现象、生物学和生态学的异速生长规律、电磁波以及复杂系统的量子进化、信号和图像处理等(参见文献[1-6]).对于大多数分数阶偏微分方程,解析解不易得到.因此,近些年来,越来越多的研究者关注其数值方法.如有限差分方法[7-8],谱方法[9-10],有限体积法[11-12],间断有限元方法[13],有限元方法[14-15]等.

本文主要考虑如下时间分数阶扩散方程

$\begin{equation}\left\{\begin{array}{ll} \displaystyle \frac{\partial^\alpha u }{\partial t^\alpha}-\Delta u=f({\bf x}, t), & ({\bf x}, t)\in\Omega\times(0, T], \\[2mm] u({\bf x}, t)=0, &({\bf x}, t)\in \partial\Omega\times(0, T], \\ { u({\bf x}, 0)}=u_0({\bf x}), & {\bf x}\in\Omega , \end{array}\right.\end{equation} $

它常被用于描述运输中的反常扩散过程[16].这里$\Omega\subset {\Bbb R}^{2}$是有界的凸多边形区域, $\partial\Omega$$\Omega$的边界, ${{\bf x}}=(x, y)$, $u_0({\bf x}), f({\bf x}, t)$为已知适当光滑的函数, $ \frac{\partial^\alpha }{\partial t^\alpha}$是关于$t$$\alpha$阶左侧Caputo导数算子,其定义为

这里$\Gamma (\cdot)$是Gamma函数.

超逼近和超收敛分析是提高有限元精度的重要分析技巧,已被广泛应用于整数阶偏微分方程的研究.如Poission方程[17],半线性伪双曲方程[18],抛物型积分微分方程[19],非线性的Sobolev方程[20],非线性的BBM方程[21].对于时间分数阶方程(1.1),文献[22]基于空间上的类Wilson元和时间上的$L1$逼近,建立了全离散格式,导出了超收敛结果.文献[23]结合空间上的两种协调和非协调混合有限元方法,构造了两种全离散逼近格式,并得到了协调元格式原始变量和中间变量的超逼近和超收敛结果,非协调元格式的最优误差估计.文献[24]探讨了多项时间扩散方程的$EQ_1^{rot}$非协调元的误差分析,得到了$L^2$模和$H^1$模的最优误差估计.在三角形单元中,线性元是自由度最少、结构最简单的,也得到了广泛应用.文献[25]探讨了椭圆方程各向异性网格下的线性三角形元的超收敛分析.文献[26-27]借助插值算子和Ritz投影算子之间的关系分别对sine-Gordon方程和Schrödinger方程导出线性三角形元的超收敛结果.

该文主要基于线性三角形元和改进的$L1$格式,构造方程(1.1)的全离散逼近格式.首先,证明了该格式在$H^1$模意义下的无条件稳定性.其次,利用该单元及Ritz投影算子的性质,导出了${L^2}$模的最优误差估计及关于投影算子在$H^1$模意义下具有$O(h^2+\tau^{2-\alpha})$阶的超逼近性质.为了得到整理超收敛结果,借助文献[26-27]中的技巧,即借助插值算子和投影算子之间的超收敛估计,进一步导出了关于插值算子具有$O(h^2+\tau^{2-\alpha})$阶的超逼近性质.从而利用插值后处理技术得到了整体超收敛估计.需要指出的是,若单独利用插值算子或者Ritz投影算子都无法得到本文的结果.最后,给出了一些数值结果验证了理论分析的有效性.

2 全离散逼近格式和一些引理

不妨令$\Omega\subset{{\Bbb R}^2}$为有界矩形区域,其边界$\partial\Omega$分别平行于$x$ -轴和$y$ -轴.假定$T_h$$\Omega$上的一族直角三角形剖分,对任意的$K\in{T_h}$,其两条边分别平行于$x$ -轴和$y$ -轴,且所有水平边和竖直边分别相等(记为GATM三角形网格剖分[27]),同时要求剖分满足正则性假设.记$h_K={\rm diam}(K), \forall K\in T_h$,即单元$K$的外接圆直径,空间步长$ h=\max\limits_{K\in T_h}{h_K}$.

定义线性三角形有限元空间

其中$v_h|_{K}$$v_h$$K$的三个顶点上的函数值唯一确定, $P_1(K)$$K$上的完全一次多项式.记$I_h: H^2(\Omega)\rightarrow V_h$为相应的线性插值算子.

问题(1.1)式的变分形式是:求$u({\bf x}, t):(0, T]\longrightarrow H_0^1(\Omega)$,满足

$\begin{equation}\left\{\begin{array}{ll}\displaystyle(\frac{\partial^\alpha u}{\partial t^\alpha}, v)+(\nabla u, \nabla v)=(f, v), \quad \forall v\in H_0^1(\Omega), \\[2mm]u(X, 0)=u_0(X).\end{array}\right.\end{equation}$

$0={t_0}<{t_1}<\cdots <{t_{N-1}}<{t_N}=T$$[0, T]$上步长为$\tau=T/N$的剖分, $t_n=n\tau, $$n=0, 1, 2, \cdots , N$.我们定义

其中$\partial_{t}\varphi^{n}=\frac{\varphi^{n}-\varphi^{n-1}}{\tau}, \tilde{b}_k=(k+1)^{1-\alpha}-k^{1-\alpha}, \ b_n^n=1, \ b_0^n=-(n^{1-\alpha}-(n-1)^{1-\alpha})$, $ b_k^n=(n-k+1)^{1-\alpha}-2(n-k)^{1-\alpha}+(n-k-1)^{1-\alpha}(1\leq k<n).$通过简单的计算不难看出$ \{\tilde{b}_k\}_0^n$是正的单调递减的序列, $b_k^n<0 (0\leq k<n)$,且$ -\sum\limits_{k=0}^{n-1}b_k^n=1$.

引入余项$ R^n=D_t^\alpha u^n-\frac{\partial^\alpha u^n}{\partial t^\alpha}$.根据文献[6]可知

$ \begin{eqnarray}\|R^n\|_0\leq C\tau^{2-\alpha}\displaystyle\max\limits_{0\leq t\leq T}\|u_{tt}({\bf x}, t)\|_{0}.\end{eqnarray}$

问题(2.1)式的全离散逼近格式为:求${U}^n\in {V}_0^h$,使得

$ \begin{equation}\left\{\begin{array}{ll}(D_t^\alpha U^n, v_h)+(\nabla U^n, \nabla v_h)=( f^n, v_h), \quad \forall v_h\in V_0^h, \\ \ U^0=R_hu_0({\bf x}), \end{array}\right.\end{equation} $

这里$R_h$$H_0^1(\Omega)\longrightarrow V_0^h$的Ritz投影算子,即对$u\in H_0^1(\Omega)$,有

$ \begin{equation}\begin{array}{cc} (\nabla(u-{R_h{u}}), \nabla v_h)=0, \forall v_h\in V_0^h. \end{array}\end{equation}$

且由文献[29]知,若$ u\in{H^1_0(\Omega)\cap H^2(\Omega)}$,该投影算子满足

$\begin{eqnarray} \|R_h{u}-u\|_0+h\|\nabla(R_h{u}-u)\|_0\leq Ch^2 \|u\|_2.\end{eqnarray}$

为了给出后面的误差分析,需要引入下列重要引理.

引理2.1[26]  设$u\in H_0^1(\Omega)\cap H^3(\Omega)$,成立

引理2.2[23]  设$\varphi^k\geq 0(k=0, 1, \cdots, L), $$\varphi^0=0, \varphi^n \leq -\sum\limits_{k=1}^{n-1} b_k^n\varphi^k +\gamma$, $\gamma>0$,则

引理2.3[23]  设$\{\xi^n\}_{n=0}^{N}$$\Omega$上的函数列,则

3 逼近格式的稳定性、超逼近、超收敛分析

本节首先讨论了逼近格式的无条件稳定性.其次证明了关于Ritz投影算子的超逼近性质.再结合投影算子和插值算子的关系得到了关于插值算子的超逼近性质.最后借助插值后处理技巧,导出了超收敛结果.

先给出如下的稳定性结果.

定理3.1  设${U}^n$是(2.3)式的解,则

其中$C_0=\max\{\Gamma(2-\alpha)\tau^\alpha, \Gamma(1-\alpha)T^\alpha\}$.

  在(2.3)式中令$v_h=\nabla D_t^\alpha{U}^n$可得

根据$D_t^\alpha{U}^n$的定义可知

由于$ -\sum\limits_{k=0}^{n-1}b_k^n\|\nabla{U}^k-\nabla{U}^n\|_0^2\geq 0$,因此

于是

$\begin{equation}\displaystyle\|\nabla{U}^n\|_0^2\displaystyle\leq -\sum\limits_{k=0}^{n-1}b_k^n\|\nabla{U}^k\|_0^2+{\Gamma(2-\alpha)}\tau^\alpha\max\limits_{t\in[0, T]}\|f({\bf x}, t)\|_0^2.\end{equation}$

接下来用数学归纳法证明如下结论

$\begin{equation}\displaystyle\|\nabla{U}^n\|_0^2\leq\|\nabla{U}^0\|_0^2+C_0\max\limits_{t\in[0, T]}\|f({\bf x}, t)\|_0^2.\end{equation}$

$n=1$时,由于$b_0^1=-1$,从(3.1)式不难得到

$\begin{equation}\displaystyle\|\nabla{U}^1\|_0^2\leq \|\nabla U^0\|_0^2+C_0\max\limits_{t\in[0, T]}\|f({\bf x}, t)\|_0^2.\end{equation}$

假设(3.2)式对于一切$0<n\leq l$都成立.由于$ \sum\limits_{k=0}^{l}b_k^{l+1}=-1$,因此利用归纳假设,当$n=l+1$时,有

这里利用了文献[22]中的结论即$\tau^\alpha\Gamma(2-\alpha)+C_0b_0^{l+1}<0$.

定理证毕.

下面再探讨超逼近性质.先分析关于Ritz投影算子的超逼近性质.为此,对任意$v_h\in {V}_0^h$,由(2.1)和(2.3)式及投影算子的性质(2.4)得到如下误差方程

$\begin{equation} \begin{array}{ll} (D_t^\alpha{\xi}^n, v_h)+(\nabla{{\xi}}^n, \nabla v_h)= -(D_t^\alpha {\eta}^n, v_h)-(R^n, v_h). \end{array} \end{equation}$

$u^n-U^n=(u^n-R_hu^n)+(R_hu^n-U^n)\doteq\eta^n+\xi^n$.

定理3.2  设$u^n, {U}^n$分别是方程(2.1)和(2.3)的解,假定$u, u_t\in H^2(\Omega), $

  在(3.4)式中令$v_h={\xi}^n$,则

根据$D_t^\alpha{\xi}^n$的定义可得

因此上式可变形为

$\begin{eqnarray}\|{\xi}^n\|_0^2+\tau^{\alpha}\Gamma(2-\alpha)\|\nabla{\xi}^n\|_{0}^{2}&=&-(\displaystyle\sum\limits_{k=0}^{n-1}b_k^n{\xi}^k, {\xi}^n)-\tau^{\alpha}\Gamma(2-\alpha)(D_t^\alpha{\eta}^n, {\xi}^n)-\tau^{\alpha}\Gamma(2-\alpha)(R^n, {\xi}^n)\nonumber\\&\doteq&\sum^{3}_{i=1}T_{i}.\end{eqnarray}$

下面分别估计$T_i\ (i=1, 2, 3)$.

由于$b_k^n <0\ (0\leq k< n)$,借助Cauchy-Schwartz不等式, $T_1$可估计为

根据$ \sum\limits_{k=0}^{n-1}\tilde{b}_k=n^{1-\alpha} (\tilde{b}_k>0)$$n\tau\leq T$$T_2$的如下估计

$\begin{eqnarray}|T_2|&\leq&\tau^{\alpha}\Gamma(2-\alpha)\|D_t^\alpha{\eta}^n\|_{0}\|{\xi}^n\|_{0}\leq\tau\|\displaystyle\sum\limits_{k=0}^{n-1}\tilde{b}_k\partial_{t}{\eta}^{n-k}\|_{0}\|{\xi}^n\|_{0}\nonumber\\&\leq&\tau\displaystyle\sum\limits_{k=0}^{n-1}\tilde{b}_k\|\partial_{t}{\eta}^{n-k}\|_{0}\|{\xi}^n\|_{0}=\tau\displaystyle\sum\limits_{k=0}^{n-1}\tilde{b}_k\|\displaystyle\frac{1}{\tau}\displaystyle\int_{t_{n-k-1}}^{t_{n-k}}{\eta}_{t}dt\|_{0}\|{\xi}^n\|_{0}\nonumber\\&\leq &C\tau h^2\displaystyle\sum\limits_{k=0}^{n-1}\tilde{b}_k\|u_{t}\|_{L^{\infty}(H^2(\Omega))}\|{\xi}^n\|_{0}\leq CT^{1-\alpha}\tau^\alphah^2\|u_{t}\|_{L^{\infty}(H^2(\Omega))}\|{\xi}^n\|_{0}, \end{eqnarray}$

其中$ \|\phi\|_{L^{\infty}(H^m(\Omega))}=\max\limits_{0\leq t\leq T} \|\phi({\bf x}, t)\|_{H^m(\Omega)}.$

利用(2.2)式有

$T_i( i=1, 2, 3)$的估计式代入(3.5)式,并且舍去正项$\tau^\alpha \Gamma(2-\alpha)\|\nabla\xi^n\|_0^2$

由引理2.2可知

$\begin{eqnarray}\|{\xi}^n\|_0\leq CT^{1-\alpha}h^2\|u_{t}\|_{L^{\infty}(H^2(\Omega))}+C\tau^{2-\alpha}\displaystyle\max\limits_{0\leq t\leq T}\|u_{tt}({\bf x}, t)\|_{0}.\end{eqnarray}$

$\|{\xi}^n\|_0$的估计式及(2.5)式不难得到如下误差估计

$\begin{eqnarray}\|u^n-{U}^n\|_{0}&\leq&\|u^n-R_h{u}^n\|_{0}+\|R_hu^n-{U}^n\|_{0}\nonumber\\&\leq &\displaystyle Ch^2(\|u\|_{L^{\infty}(H^2(\Omega))}+\|u_{t}\|_{L^{\infty}(H^2(\Omega))})+C\tau^{2-\alpha}\max\limits_{0\leq t\leq T}\|u_{tt}({\bf x}, t)\|_{0}.\end{eqnarray}$

接下来再来估计${\xi}^n$$H^1$模意义下的误差.为此,在(3.4)式中令$v_h=D_t^\alpha{\xi}^n $

基于$D_t^\alpha{\xi}^n$的定义及引理2.3,有

$\begin{eqnarray}(\nabla{\xi}^n, \nabla D_t^\alpha{\xi}^n)&=&\frac{\tau^{-\alpha}}{\Gamma(2-\alpha)}(\nabla{\xi}^n, \sum\limits_{k=0}^nb_k^n\nabla{\xi}^k)\nonumber\\ &=&\displaystyle\frac{\tau^{-\alpha}}{2\Gamma(2-\alpha)}(\|\nabla{\xi}^n\|_0^2+ \displaystyle\sum\limits_{k=0}^{n-1}b_k^n\|\nabla{\xi}^k\|_0^2-\displaystyle\sum\limits_{k=0}^{n-1}b_k^n\|\nabla{\xi}^k-\nabla{\xi}^n\|_0^2). \end{eqnarray}$

类似于(3.6)式中$\|D_t^\alpha {\eta}^n\|_0$的估计及Young不等式可得

$\begin{equation}|(D_t^\alpha {\eta}^n, D_t^\alpha{\xi}^n)|\leq\displaystyle\frac{1}{2}\|D_t^\alpha{\eta}^n\|_0^2+\displaystyle\frac{1}{2}\|D_t^\alpha{\xi}^n\|_0^2 \leqCh^4\|u_t\|_{L^{\infty}(H^2(\Omega))}^2+\displaystyle\frac{1}{2}\|D_t^\alpha{\xi}^n\|_0^2.\end{equation}$

基于(2.2)式得

$\begin{equation}\displaystyle|-(R^n, D_t^\alpha{\xi}^n)|\leq\|R^n\|_0\|D_t^\alpha{\xi}^n\|_0\leqC\tau^{4-2\alpha}\max\limits_{0\leq t\leq T}\|u_{tt}({\bf x}, t)\|_{0}^2+\displaystyle\frac{1}{2}\|D_t^\alpha{\xi}^n\|_0^2.\end{equation}$

综合(3.9)-(3.11)式和$ -\sum\limits_{k=0}^{n-1}b_k^n\|\nabla{\xi}^k-\nabla{\xi}^n\|_0^2>0$

由引理2.2可知

$\begin{eqnarray}\|\nabla{\xi}^n\|_0^2\leq Ch^4\|u_t\|_{L^{\infty}(H^2(\Omega))}^2+C\tau^{4-2\alpha}\max\limits_{0\leq t\leq T}\|u_{tt}({\bf x}, t)\|_{0}^2.\end{eqnarray}$

结合(3.7)和(3.12)式,得定理3.2的第二式.定理证毕.

采用文献[26-27]中的方法,借助投影算子、插值算子的性质导出二者的超收敛估计,再结合定理3.2导出了关于插值算子的超逼近结果.

定理3.3  设$u^n, {U}^n$分别是方程(2.1)和(2.3)的解,假定$u\in H^3(\Omega), u_t\in H^2(\Omega), $

  由于$(\nabla (R_hu^n-u^n), \nabla(U^n-I_hu^n))=0$,则

根据定理3.2有

利用引理2.1可得

因此

定理证毕.

为了进行整体超收敛分析,我们可以根据文献[28]在$\overline{K}$上构造插值后处理算子${\Pi}_{2h}$:

这里$P_2(\overline{K})$$C(\overline{K})$分别为$\overline{K}$上完全二次多项式和连续函数空间, $\overline{K}\in T_{2h}$是由相邻4个$T_h$中的单元合并构成一个大单元.插值算子${\Pi}_{2h}$满足

$\begin{equation}{\Pi}_{2h} I_h{u}={\Pi}_{2h}{u}, \end{equation}$

$\begin{equation}\|{\Pi}_{2h}u-u\|_1\leq Ch^2\|u\|_3, \ \ u\in H^3(\Omega), \end{equation}$

$\begin{equation}\|{\Pi}_{2h} v\|_1\leq C\|v\|_1, \ \ \ \forall v\in S_2^h, \end{equation}$

其中$S_2^h$为完全二次有限元空间.

应用定理3.3和插值算子${\Pi}_{2h}$的性质(3.13)-(3.15)式不难得到如下超收敛结果.

定理3.4  在定理3.3的条件下有

注3.1  在定理3.2的讨论中,如果将投影算子$R_h$变成插值算子$I_h$,误差方程(3.4)中右端多了一项$ -(\nabla\eta^n, \nabla v_h)$.根据文献[28]知

(1)此时最优误差为

(2)在估计$\|\xi^n\|_1$时,利用逆估计$ -(\nabla\eta^n, \nabla v_h)\leq Ch\|u^n\|_3\|v_h\|_0$,这时只能得到

因此,可以看出虽然单独利用插值算子$I_h$得到了$L^2$模意义下的最优误差估计,但解的光滑度需要从$H^2(\Omega)$提高到$H^3(\Omega)$,且此时得不到在$H^1$模意义下的超逼近结果.

4 数值结果

本节为了验证理论分析的有效性和正确性,给出了一些数值结果.

考虑如下的时间分数阶扩散方程

这里$\Omega=(-1, 1)\times(-1, 1), \ T=[0, 1]$, $f({\bf x}, t)=\frac{t^{1-\alpha}}{\Gamma(2-\alpha)}(x^2-1)(y^2-1)-2t(x^2+y^2-2)$, $u_0({\bf x})=u({\bf x}, 0)=0.$该方程的解是$u({\bf x}, t)=t(x^2-1)(y^2-1)$.

表 1-2中,通过选取$h\approx\tau^{2-\alpha}$$h^2\approx\tau^{2-\alpha}$,分别给出当$\alpha=0.85, 0.15$时取$t_n=0.1, $$ 0.01$$\|u^n-U^n\|_{0}$, $\|I_hu^n-U^n\|_{1}$$\|u^n-\Pi_{2h}^2U^n\|_{1}$在时间方向的误差估计.不难看出在时间方向$\|u^n-U^n\|_{0}$, $\|I_hu^n-U^n\|_{1}$$\|u^n-\Pi_{2h}U^n\|_{1} $的收敛阶接近于$2-\alpha$,与我们的理论分析相吻合.

表 1   $\alpha=0.85$时在$t_n=0.1, t_n=0.01$处时间方向的误差和收敛阶

$ t_n$$\tau$$\|u^n-U^n\|_{0}$$\|I_hu^n-U^n\|_{1}$$\|u^n-\Pi_{2h}U^n\|_{1}$
${t_n}/{2}$3.869e-03/4.026e-03/2.121e-02/
${t_n}/{4}$1.687e-031.1981.780e-031.1789.538e-031.153
0.1${t_n}/{8}$7.486e-041.1727.959e-041.1614.292e-031.152
${t_n}/{16}$3.368e-041.1523.570e-041.1571.934e-031.150
${t_n}/{2}$2.091e-05/1.294e-05/1.656e-04/
${t_n}/{4}$9.240e-061.1785.780e-061.1637.369e-051.168
0.01${t_n}/{8}$4.125e-061.1642.596e-061.1553.308e-051.156
${t_n}/{16}$1.855e-061.1531.169e-061.1511.490e-051.150

新窗口打开| 下载CSV


表 2   $\alpha=0.15$时在$t_n=0.1, t_n=0.01$处时间方向的误差和收敛阶

$ t_n$$\tau$$\|u^n-U^n\|_{0}$$\|I_hu^n-U^n\|_{1}$$\|u^n-\Pi_{2h}U^n\|_{1}$
${t_n}/{2}$6.913e-04/9.330e-04/2.740e-03/
${t_n}/{4}$1.890e-041.8712.537e-041.8797.466e-041.876
0.1${t_n}/{8}$5.174e-051.8696.974e-051.8632.044e-041.869
${t_n}/{16}$1.430e-051.8561.9267e-051.8565.649e-051.855
${t_n}/{2}$6.845e-05/8.960e-05/2.832e-04/
${t_n}/{4}$1.881e-051.8642.461e-051.8647.790e-051.862
0.01${t_n}/{8}$5.196e-061.8566.799e-061.8562.153e-051.855
${t_n}/{16}$1.440e-061.8511.885e-061.8515.968e-061.851

新窗口打开| 下载CSV


表 3-5中,给出了当选取不同的$\alpha$,即$\alpha=0.3, 0.6, 0.9$和时间步长$\tau=0.001$时, $\|u^n-U^n\|_{0}$, $ \|I_h u^n-U^n\|_{1}$$\|u^n-\Pi_{2h}U^n\|_{1}$在不同时间点$t_n=0.2, 0.4, 0.7, 0.9$处的误差和收敛阶,结果表明逼近阶和收敛阶接近于二阶.因此数值结果表明超逼近与超收敛与我们的理论分析是一致的.

表 3   $\|u^n-U^n\|_{0}$取不同$\alpha$时在空间方向的误差和收敛阶

$\alpha$$h$$t_{n}=0.2$$t_{n}=0.4$$t_{n}=0.7$$t_{n}=0.9$
1/43.952e-02/8.084e-02/1.437e-01/1.860e-01/
$1/8$1.042e-021.9232.140e-021.9183.814e-021.9144.942e-021.912
0.3$1/16$2.640e-031.9815.424e-031.9809.676e-031.9791.254e-021.979
$1/32$6.622e-041.9951.361e-031.9952.428e-031.9953.146e-031.995
1/43.642e-02/7.756e-02/1.413e-01/1.844e-01/
1/89.487e-021.9412.039e-021.9283.737e-021.9194.889e-021.915
0.6$1/16$2.395e-031.9865.158e-031.9839.471e-031.9801.240e-021.979
$1/32$6.002e-041.9971.293e-031.9952.376e-031.9953.111e-031.995
$1/4$3.229e-02/7.331e-02/1.388e-01/1.833e-01/
$1/8$8.277e-031.9641.908e-021.9423.656e-021.9254.851e-021.918
0.9$1/16$2.081e-031.9924.814e-031.9869.255e-031.9821.230e-021.980
$1/32$5.209e-041.9981.206e-031.9972.321e-031.9963.084e-031.995

新窗口打开| 下载CSV


表 4   $\|U^n-I_hu^n\|_{1}$取不同$\alpha$时在空间方向的误差和收敛阶

$\alpha$$h$$t_{n}=0.2$$t_{n}=0.4$$t_{n}=0.7$$t_{n}=0.9$
$1/4$5.310e-02/1.103e-01/1.982e-01/2.575e-01/
$1/8$1.387e-021.9372.894e-021.9305.217e-021.9256.790e-021.923
0.3$1/16$3.506e-031.9847.324e-031.9821.321e-021.9811.720e-021.981
$1/32$8.791e-041.9951.837e-031.9963.315e-031.9954.316e-031.995
$1/4$4.606e-02/1.028e-01/1.926e-01/2.538e-01/
$1/8$1.183e-021.9612.673e-021.9445.048e-021.9326.676e-021.927
0.6$1/16$2.977e-031.9906.749e-031.9861.277e-021.9831.690e-021.982
$1/32$7.457e-041.9971.691e-031.9963.202e-031.9964.240e-031.995
$1/4$3.674e-02/9.320e-02/1.870e-01/2.514e-01/
$1/8$9.223e-031.9942.389e-021.9644.872e-021.9406.593e-021.931
0.9$1/16$2.311e-031.9976.012e-031.9911.231e-021.9851.668e-021.983
$1/32$5.781e-041.9991.505e-031.9983.085e-031.9964.182e-031.996

新窗口打开| 下载CSV


表 5   $\|u^n-\Pi_{2h}U^n\|_{1}$取不同$\alpha$时在空间方向的误差和收敛阶

$\alpha$$h$$t_{n}=0.2$$t_{n}=0.4$$t_{n}=0.7$$t_{n}=0.9$
$1/4$1.540e-01/3.080e-01/5.390e-01/6.931e-01/
$1/8$4.179e-021.8828.357e-021.8821.462e-011.8821.880e-011.882
0.3$1/16$1.067e-021.9702.133e-021.9703.733e-021.9704.800e-021.970
$1/32$2.681e-031.9925.361e-031.9929.383e-031.9921.206e-021.992
$1/4$1.545e-01/3.082e-01/5.391e-01/6.931e-01/
$1/8$4.190e-021.8828.362e-021.8821.462e-011.8821.880e-011.882
0.6$1/16$1.069e-021.9702.134e-021.9703.733e-021.9704.780e-021.970
$1/32$2.687e-031.9935.364e-031.9929.382e-031.9921.206e-021.992
$1/4$1.557e-01/3.088e-01/5.391e-01/6.931e-01/
$1/8$4.221e-021.8838.379e-021.8821.463e-011.8821.880e-011.882
0.9$1/16$1.077e-021.9712.138e-021.9703.734e-021.9704.800e-021.970
$1/32$2.071e-031.9935.374e-031.9939.384e-031.9921.206e-021.992

新窗口打开| 下载CSV


参考文献

He J H. Nonlinear Oscillation with Fractional Derivative and its Applications//Wen B C. International Conference on Vibrating Engineering' 98. Shenyang:Northeastern Univ Press, 1998:288-291

[本文引用: 1]

Ming C , Liu F , Zheng L , et al.

Analytical solutions of multi-term time fractional differential equations and application to unsteady flows of generalized viscoelastic fluid

Computational Methods in Applied Mathematics, 2016, 72: 2084- 2097

URL    

He J H .

Approximate analytical solution for seepage flow with fractional derivatives in porous media

Computer Methods in Applied Mechanics & Engineering, 1998, 167 (1-2): 57- 68

URL    

Mainardi F. Fractional Calculus, Some Basic Problems in Continuumand Statisticalmechanics//Carpinteri A, Mainardi F, et al. Fractals and Fractional Calculus in Continuum Mechanics. New York:Springer Verlag, 1997:291-348

Kilbas A, Srivastava H, Trujillo J. Theory and Applications of Fractional Differential Equations. Amsterdam:Elsevier, 2006

Liu F , Zhuang P , Liu Q . Numerical Methods of Fractional Partial Differential Equations and Applications. Beijing: Science Press, 2015

[本文引用: 2]

Zhang Y N , Sun Z Z , Liao H L .

Finite difference methods for the time fractional diffusion equations and non-uniform meshes

Journal of Computational Physics, 2014, 265 (3): 195- 210

URL     [本文引用: 1]

Ye H , Liu F , Anh V .

Compact difference scheme for distributed-order time-fractional diffusion-wave equation on bounded domains

Journal of Computational Physics, 2015, 98: 652- 660

URL     [本文引用: 1]

Zeng F , Liu F , Li C , et al.

Crank-nicolson ADI spectral method for the two-dimensional Riesz space fractional nonlinear reaction-diffusion equation

SIAM Journal on Numerical Analysis, 2014, 52 (6): 2599- 2622

DOI:10.1137/130934192      [本文引用: 1]

Zheng M , Liu F , Anh V , et al.

A high-order spectral method for the multi-term time-fractional diffusion equations

Applied Mathematical Modelling, 2016, 40 (7-8): 4970- 4985

DOI:10.1016/j.apm.2015.12.011      [本文引用: 1]

Liu F , Zhuang P , Turner I , et al.

A new fractional finite volume method for solving the fractional diffusion equation

Applied Mathematical Modelling, 2014, 38 (15-16): 3871- 3878

DOI:10.1016/j.apm.2013.10.007      [本文引用: 1]

Jia J , Wang H .

A fast finite volume method for conservative space-fractional diffusion equations in convex domains

Journal of Computational Physics, 2016, 310: 63- 84

DOI:10.1016/j.jcp.2016.01.015      [本文引用: 1]

Wei L .

Analysis of a new finite difference/local discontinuous Galerkin method for the fractional diffusionwave equation

Numerical Algorithms, 2017, 304: 180- 189

URL     [本文引用: 1]

Jin B , Lazarov R , Liu Y , et al.

The Galerkin finite element method for a multi-term time-fractional diffusion equation

Journal of Computational Physics, 2015, 281: 825- 843

DOI:10.1016/j.jcp.2014.10.051      [本文引用: 1]

Zhuang P , Liu F , Turner I , et al.

Galerkin finite element method and error analysis for the fractional cable equation

Numerical Algorithms, 2016, 72 (2): 447- 466

DOI:10.1007/s11075-015-0055-x      [本文引用: 1]

Metzler R , Klafter J .

The randomwalk's guide to anomalous diffusion:a fractional dynamics approach

Physics Reports, 2000, 339 (1): 1- 77

DOI:10.1016/S0370-1573(00)00070-3      [本文引用: 1]

Lin Q , Tobiska L , Zhou A H .

Superconvergence and extrapolation on nonconforming low order finite elements applied to the Poission equation

IMA Journal of Numerical Analysis, 2005, 25 (1): 160- 181

URL     [本文引用: 1]

石东洋, 史艳华.

半线性伪双曲方程最低阶的H1-Galerkin混合元方法

系统科学与数学, 2015, 35 (5): 514- 526

URL     [本文引用: 1]

Shi D Y , Shi Y H .

The lowest order H1-Galerkin mixed finite element method for semi-linear pseudohyperbolic equation

Journal of Systems Science and Mathematical Sciences, 2015, 35 (5): 514- 526

URL     [本文引用: 1]

张铁.

抛物型积分-微分方程有限元近似的超收敛性质

高等学校计算数学学报, 2001, 23 (3): 193- 201

DOI:10.3969/j.issn.1000-081X.2001.03.001      [本文引用: 1]

Zhang T .

Superconvergence of finite element approximations to integro-differential equations of parabolic type

Numerical Mathematies A Journal of Chinese Universities, 2001, 23 (3): 193- 201

DOI:10.3969/j.issn.1000-081X.2001.03.001      [本文引用: 1]

Shi D Y , Wang J J .

Superconvergence analysis of an H1-Galerkin mixed finite element method for Sobolev equations

Computers & Mathematics with Applications, 2016, 72 (6): 1590- 1602

URL     [本文引用: 1]

Shi D Y , Yang H J .

A new approach of superconvergence analysis for nonlinear BBM equation on anisotropic meshes

Applied Mathematics Letters, 2016, 58: 74- 80

DOI:10.1016/j.aml.2016.02.007      [本文引用: 1]

Zhao Y M , Zhang Y D , Shi D Y , et al.

Superconvergence analysis of nonconforming fnite element method for two-dimensional time fractional difusion equations

Applied Mathematics Letters, 2016, 59: 38- 47

DOI:10.1016/j.aml.2016.03.005      [本文引用: 2]

Zhao Y M , Chen P , Bu W P , et al.

Two mixed finite element methods for time-fractional diffusion equations

Journal of Scientific Computing, 2017, 70 (1): 407- 428

DOI:10.1007/s10915-015-0152-y      [本文引用: 3]

Zhao Y M , Zhang Y D , Liu F , et al.

Analytical solution and nonconforming finite element approximation for the 2D multi-term fractional subdiffusion equation

Applied Mathematical Modelling, 2016, 40: 8810- 8825

DOI:10.1016/j.apm.2016.05.039      [本文引用: 1]

石东洋, 梁慧.

各向异性网格下线性三角形元的超收敛性分析

工程数学学报, 2007, 24 (3): 487- 493

DOI:10.3969/j.issn.1005-3085.2007.03.014      [本文引用: 1]

Shi D Y , Liang H .

The superconvergence analysis of linear triangular element on anisotropic meshes

Chinese Journal of Engineering Mathematics, 2007, 24 (3): 487- 493

DOI:10.3969/j.issn.1005-3085.2007.03.014      [本文引用: 1]

石东洋, 王芬玲, 赵艳敏.

非线性sine-Gordon方程的各向异性线性元高精度分析新模式

计算数学, 2014, 36 (3): 245- 256

URL     [本文引用: 4]

Shi D Y , Wang F L , Zhao Y M .

A new pattern of high accuracy analysis of anisotropic linear element for nonlinear sine-gordon equations

Mathematica Numerica Sinica, 2014, 36 (3): 245- 256

URL     [本文引用: 4]

Shi D Y , Wang P L , Zhao Y M .

Superconvergence analysis of anisotropic linear triangular finite element for nonlinear Schrödinger equation

Applied Mathematics Letters, 2014, 38 (38): 129- 134

URL     [本文引用: 4]

林群, 严宁宁. 高效有限元构造与分析. 保定: 河北大学出版社, 1996

[本文引用: 2]

Lin Q , Y N N . The Construction and Analysis of High Eficient Finite Element Methods. Baoding: Hebei University Press, 1996

[本文引用: 2]

张铁. 偏微分-积分方程的有限元方法. 北京: 科学出版社, 2009

[本文引用: 1]

Zhang T . Finite Element Methods for Partial Differentio-Integral Equations. Beijing: Science Press, 2009

[本文引用: 1]

/