Message Board

Dear readers, authors and reviewers,you can add a message on this page. We will reply to you as soon as possible!

2019 Volume 44 Issue 11
Article Contents

Hong-yan SHEN, Ming LI. Convergence Analysis of a Cascadic Multigrid Algorithm Combined with Quadratic Finite Element Method[J]. Journal of Southwest China Normal University(Natural Science Edition), 2019, 44(11): 24-28. doi: 10.13718/j.cnki.xsxb.2019.11.004
Citation: Hong-yan SHEN, Ming LI. Convergence Analysis of a Cascadic Multigrid Algorithm Combined with Quadratic Finite Element Method[J]. Journal of Southwest China Normal University(Natural Science Edition), 2019, 44(11): 24-28. doi: 10.13718/j.cnki.xsxb.2019.11.004

Convergence Analysis of a Cascadic Multigrid Algorithm Combined with Quadratic Finite Element Method

More Information
  • Corresponding author: Ming LI
  • Received Date: 19/09/2017
    Available Online: 20/11/2019
  • MSC: O241

  • A new cascadic multigrid (SECMG) algorithm combined with quadratic finite element method has been proposed, in which a quadratic interpolation operator has been constructed by means of the nodes information of quadratic finite element cell, to provide an initial guess of iterative solution on the next fine grid. The convergence of the developed algorithm has been analyzed. Numerical results have been presented to verify the feasibility of this new method.
  • 加载中
  • [1] BORNEMANN F A, DEUFLHARD P.The Cascadic Multigrid Method for Elliptic Problems[J].Numerische Mathematik, 1996, 75(2): 135-152.

    Google Scholar

    [2] 石钟慈, 许学军, 黄云清.经济的瀑布型多重网格法(ECMG)[J].中国科学(A辑), 2007, 37(9):1083-1098.

    Google Scholar

    [3] SHI Z C, XU X J.Cascadic Multigrid Method for Elliptic Problems[J].East-West Journal of Numerical Mathematics, 1999, 7(3): 199-209.

    Google Scholar

    [4] 石钟慈, 许学军.Cascadic Multigrid for Parabolic Problems[J].计算数学(英文版), 2000, 18(5):551-560.

    Google Scholar

    [5] 石钟慈, 许学军.一类新的瀑布型多重网格法[J].中国科学(A辑), 2000, 30(9):799-807.

    Google Scholar

    [6] 李郴良, 陈传淼, 许学军.基于超收敛和外推方法的一类新的瀑布型多重网格方法[J].计算数学, 2007, 29(4):439-448. doi: 10.3321/j.issn:0254-7791.2007.04.011

    CrossRef Google Scholar

    [7] CHEN C M, HU H L.Extrapolation Cascadic Multigrid Method on Piecewise Uniform Grid[J].Science China Mathematics, 2013, 56(12): 2711-2722. doi: 10.1007/s11425-013-4732-8

    CrossRef Google Scholar

    [8] CHEN C M, HU H L, XIE Z Q, et al.Analysis of Extraplolation Cascadic Multigrid Method (EXCMG)[J]. Science in China, 2008, 51(8): 1349-1360. doi: 10.1007/s11425-008-0119-7

    CrossRef Google Scholar

    [9] LI M, LI C L, CUI X Z, et al.Cascadic Multigrid Methods Combined with Sixth Order Compact Scheme for Poisson Equation[J].Numerical Algorithms, 2016, 71(4): 715-727. doi: 10.1007/s11075-015-0018-2

    CrossRef Google Scholar

    [10] 王烈衡, 许学军.有限元方法的数学基础[M].北京:科学出版社, 2004.

    Google Scholar

    [11] BANK R, XU J C.Asymptotically Exact A Posteriori Error Estimators, Part Ⅰ:Grids with Super Convergence[J].SIAM Journal on Numerical Analysis, 2003, 41(6):2294-2312. doi: 10.1137/S003614290139874X

    CrossRef Google Scholar

    [12] 陈传淼, 黄云清.有限元高精度理论[M].长沙:湖南科学技术出版社, 1995

    Google Scholar

    [13] XU J, ZHANG Z.Analysis of Recovery Type A Posteriori Error Estimators for Mildly Structured Grids[J].Mathematics of Computation, 2004, 73:1139-1152.

    Google Scholar

    [14] 石钟慈, 王鸣.有限元方法[M].北京:科学出版社, 2010.

    Google Scholar

    [15] 李明, 赵金娥.二维椭圆问题的经济外推瀑布多重网格法[J].西南大学学报(自然科学版), 2014, 36(7):68-72.

    Google Scholar

    [16] 李明, 崔向照, 赵金娥.求解高次有限元方程的外推瀑布型多重网格法[J].西南师范大学学报(自然科学版), 2016, 41(1):20-23.

    Google Scholar

    [17] WANG B, MENG F, FANG Y.Efficient Implementation of RKN-type Fourier Collocation Methods for Second-Order Differential Equations[J].Applied Numerical Mathematics, 2017, 119:164-178. doi: 10.1016/j.apnum.2017.04.008

    CrossRef Google Scholar

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Tables(2)

Article Metrics

Article views(817) PDF downloads(137) Cited by(0)

Access History

Other Articles By Authors

Convergence Analysis of a Cascadic Multigrid Algorithm Combined with Quadratic Finite Element Method

    Corresponding author: Ming LI

Abstract: A new cascadic multigrid (SECMG) algorithm combined with quadratic finite element method has been proposed, in which a quadratic interpolation operator has been constructed by means of the nodes information of quadratic finite element cell, to provide an initial guess of iterative solution on the next fine grid. The convergence of the developed algorithm has been analyzed. Numerical results have been presented to verify the feasibility of this new method.

  • 高次有限元法对偏微分方程(PDE)问题的逼近较好,已得到广泛应用.多重网格(MG)法[1-9]是求解以偏微分方程为背景的离散化方程的有效数值算法之一.为进一步讨论求解二次有限元方程的瀑布型多重网格法,本文使用二次插值作为插值算子,针对二次有限元方程,提出了改进的瀑布型多重网格法(SECMG),讨论了该算法的收敛性.数值实验表明,改进算法具有较好的计算效果.本文的思想和算法可推广到其他高次元.

1.   模型问题
  • 考虑带Dirichlet边界条件的二维椭圆型偏微分方程:

    其中α(xy),β(xy),f(xy)为给定的函数且满足∀(x0y0)∈Ωα(x0y0)>0,β(x0y0)≥0.如果α(xy)≡1,β(xy)≡0,模型(1)为泊松方程;如果α(xy)≡1,β(xy)≡f(xy)≡0,模型(1)为拉普拉斯方程.

    使用一致网格剖分Ω,可得不同步长$ h_{j}=\frac{h_{M}}{2^{M-j}}(j=1, 2, \cdots, M)$下的一系列离散化网格Ωj,易知Ω1ΩM分别为最细、最粗网格层.使用二次有限元在网格Ωj上离散模型问题(1),可得对应的离散化方程组

    其中:Aj为对称正定稀疏矩阵,Fj为右端列向量,uj为未知解向量.

2.   改进的瀑布型多重网格法
  • 文献[6]提出,在线性有限元离散网格中,将相邻的几个单元组合成一个大单元,进而构造一个二次插值算子,用于给相邻细层提供一个好的初始值,能在一定程度上加快瀑布型多重网格法的收敛速度.借鉴文献[6]的思想,直接使用每个二次三角形单元的6个节点(3个顶点、3个中点)的信息,无需合并相邻单元,即可构造网格Ωj上的基于每个单元的二次插值算子Ψj2,且满足

    1) ‖v-Ψj2v‖≤Chjr-l+1vr+1,1≤r≤2,vHr+1(Ω),l=0,1.

    2) ‖v-Ψj2vlcvlvVjl=0,1,j=1,2,…,M.

    类似于文献[6]的思路,使用二次插值算子Ψj2为相邻细层网格提供初始值,可构造求解方程组(2)的瀑布型多重网格法(SECMG).

    算法1  瀑布型多重网格法(SECMG)

    步骤1  精确求解最粗网格层方程AMuM=FM,得uM*

    步骤2  对j=M-1,…,2,1

    (a) 插值得到第j层的初始值uj0:=Ψj+12uj+1*

    (b) 对uj0使用CG(共轭迭代法)磨光mj次,ujmj:=CG(uj0mj);

    (c) 令uj*:=ujmj.

    其中mj表示第j层网格上的磨光步数[2].

3.   收敛性分析
  • 在讨论算法1的收敛性之前,先引入如下引理.

    引理1[10]  对任意vVj,有

    引理2[6]  令CjVjVj表示在第j层上的迭代算子,并定义线性算子KjVjVj,使得

    算子Kj具有如下性质:

    其中:mj表示第j层的迭代次数,r是与所采用的迭代方法有关的常数.本文采用共轭梯度法(CG)迭代,对应于r=1.

    引理3[11-13]  设u为问题(1)的解,ujIj分别为u在第j层上的有限元解和有限元插值,有以下超收敛性:

    引理4[14]  设Vh,2是二次三角形单纯形元所对应的有限元空间,则对任意fL2(Ω),问题(1)有唯一解uh,并且$ \mathop {\lim }\limits_{h \to 0} {\left\| {\mathit{\boldsymbol{u}} - {\mathit{\boldsymbol{u}}_h}} \right\|_{1, \mathit{\Omega }}} = 0$.当uH01(Ω)∩H3(Ω)时,有

    引理5  在引理4的条件上,有

      由‖u-uh0≤‖u-uh1,|u|3≤‖u3,结合引理4即可证得‖u-uh0,Ωh2u3,Ω.证毕.

    在上述引理的基础上,给出如下定理及证明过程.

    定理1  若uH01(Ω)∩H3(Ω),对算法1有

    下面对(3)式展开讨论.

    和引理5可得

    由引理2可得

    综上可得

    利用递归关系

    可得

    证毕.

4.   数值实验
  • 为了验证本文算法的有效性,给出如下数值算例.

    例1  考虑区域Ω:[0, 1]×[0, 1]上的泊松方程

    其Dirichlet边界条件和右端源函数f(xy)取决于真解

    例2  考虑区域Ω:[0, 1]×[0, 1]上的带变系数椭圆型方程

    其中α(xy)=exy+2,β(xy)=1,Dirichlet边界条件和右端源函数f(xy)取决于真解

    为便于比较,记通常的瀑布型多重网格法为CMG,其插值算子为线性插值,磨光算子为CG迭代.取M=2,并约定G表示最细层网格规模;I表示最细层上的磨光步数;E表示问题真解u与算法求得的近似解u1*的能量范数误差|||u-u1*|||;t表示计算时间.算例的数值结果见表 12.

    表 12可以看出,与算法(CMG)相比,SECMG法虽然在计算时间方面要稍多些,但在计算精度方面具有一定的优势,计算精度要高一个量级,验证了本文算法的有效性.

Table (2) Reference (17)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return