南京大学学报(自然科学版) ›› 2012, Vol. 48 ›› Issue (6): 753–760.

• • 上一篇    下一篇

 区域分解预处理器研究及其在地下水模拟中的应用*

 王佩1,朱国荣1**,江思氓2,孙斌堂3   

  • 出版日期:2015-09-10 发布日期:2015-09-10
  • 作者简介: (1.南京大学水科学系,南京,210093;2.同济大学水利工程系,上海,200092
    3.中国冶金地质工程勘察总局山东局,济南,250014
  • 基金资助:
     国家自然科学基金(41002078)

 Domain decomposition preprocessor research and
its application in groundwater modeling

 Wang Pei1,Zhu Uuo-Rong1 ,Jung Si-Min2,Sun Bin-Tang3   

  • Online:2015-09-10 Published:2015-09-10
  • About author: (1 .Department of Hydrosciences,Nanjing University,Nanjing,210093,China;2. Department of Hydraulic
    Engineering,Tongji University,Shanghai,200092,China;3. Shandong Bureau of
    China Exploration and Engineering Bureau,Jinan,250014,China)

摘要:  地下水数值计算中,如何提高大型稀疏线性代数方程组的求解效率一直备受关注.条件预处
理共辆梯度方法是求解大型稀疏线性代数方程组的有效方法,而如何构造高效的预处理器至关重要.木
文基于有限元理论介绍了区域分解预处理器(domain decomposition preprocessor,DDP)的设计原理及
实现过程,将其与预处理共辆梯度法结合为区域分解预处理共辆梯度法(DDP-PCG)并应用于地下水模
拟中.文中首先采用具有解析解的均质稳定流模型验证了DDP-PCG的可靠性;接着对该模型研究区进
行不同规模网格剖分,均采用CG, Jacobi-PCG, SSOR-PCG, DDP-PCG求解,结果表明DDP-PCG迭代收
敛次数几乎不随网格规模发生变化,具有很强的鲁棒性;随着网格规模的增加,DDP-PCG求解效率优势
更加显著.最后针对承压含水层介质参数连续变化和突变两种非均质问题,对研究区进行大规模网格剖
分,进一步证明在相同误差限下,DDP-PCG的求解效率高于其它三种方法.

Abstract:  The problem that how to improve the efficiency of solving the large sparse linear algebraic equations has
always been of concern in groundwater numerical simulation. Preconditioned conjugate gradient method is an effective
method for solving large sparse linear algebraic equations. And how to construct efficient preprocessor is essential.
The domain decomposition preprocessor design principles and implementation process are descirbed based on the ti-
nite clement theory.The preconditioned conjugate gradient method with domain decomposition preconditioner(DDP-
PCU)is proposed and applied to groundwater simulation. Firstly, DDP-PCU’s reliability is proved with analytical
homogeneous model. Then CU,Jacobi-PCU,SSOR-PCU,DDP-PCU are used to solve the model under different subdi-
visions,which show that DDP-PCU is of strong robustness and the higher the subdivision is, the more notable is
DDP-PCG’s efficiency. Finally, inhomogencous confined aquifer with continuous coefficients and abrupt coefficients
are caculated under finely subdivision,which further prove that under the same error limit,DDP-PCU’s efficiency are
the highest of all.

[1]Lv T,Shi J M, Lin Z B. Domain decomposition methods-new numerical techniques for solving PDE. Beijing;Science Press, 1999.(吕涛,石济民,林振宝.区域分解算法-扁微分方程数值解新技术.北京:科学出版社,1999).
[2]Zhang Y J,Sun Q. Preconditioned matrix and its structure methods. Journal of Changchun University of science and technology, 2006,29(4):128一130. (张永杰,孙秦.预处理矩阵及其构造方法.长春理工大学学报,2006,29(4):128-130).
[3]WuXP,Xu G M, incomplete cholesky conjugate gradient method and its applation in geoelectric field computation,OGP,1998,33(1):89一9’1.(吴小平,徐果明.不完全Cholcsky共辆梯度法及其在地电场计算中的应用.石油地球物理勘探,1998,33(1):89一94).
[4]Tan X X, Xi G. Applation of conjugate gradient  method in three-dimensional complex flow. Journal of Hydrodynamics(A),2002,17(1):116一123.(谭欣星,席光.共辆梯度法在三维复杂流动数值分析中的应用.水动力学研究与进展(A),2002,17 (1):116一123).
[5]Bao J Q Yang Q, Chen Y R, et al. Application of symmetric successive over relaxation precondition conjugate method to largrscale clasto-plastic FEM analusis for high arch dams. Journal of Hydraulic Engineering, 2009 , 40 ( 5 ) ; 589 -595.(包劲青,杨强,陈英儒等.对称超松弛预处理共辆梯度方法在高拱坝整体大规模弹塑性有限元分析中的应用. 水利学报,2009,40(5):589-595).
[6]Zhang Y J,Sun Q, Li J H. An improved ICCG method for large scale sparse linear equations. Chinese Journal of Computational Physics,2007,24(5):581-584.(张永杰,孙秦,李江海.大型稀疏线性方程组的改进ICCG方法.计算物理,2007,21(5):581一584).
[7]Chan T F. Analysis of prcconditioncrs for domain decomposition. SIAM Journal on Numerical Anal- ysis. 1987.24(2) .382一390.
[8]Shu J W , gui L Z , Zhou W S , et al. Parallel com putmg of domain decomposition methods for sol  ving numerical simulation of black oil resenvoir. Journal of Nanjing University(Natural sciences), 1999,35(1):51-57.舒继武,归丽忠,周维四
等.区域分解法解黑油数值模拟问题的并行计算.南京大学学报(自然科学),1999,35(1):51一57).
[9]Liu Q K,Uui I. Z,Shu J W,et al. improved parallel computing of domain decomposition methods for sol- wing numerical simulation of black oil reservoir. Jour- nal of Nanjing University(Natural Sciences),2003,39(2);229-237.(刘青昆,归丽忠,舒继武等.区
域分解法解黑油数值模拟问题改进的并行计算. 南京大学学报(自然科学),2003,39(2);229- 237).
[10]Bramble J H,Pasciak J E,Schatz A H. An itcra tive method for elliptic problems on regions parti boned into substructures. Mathematic of Compu tation, 1986,174 (46):361一369.
[11]Bramble J H,Pasciak J E,Schatz A H.The con- struction of preconditioner for elliptic problrms by substructuring I.mathcmatic of computa- tion, 1986,175(47):103一134.
[12]Bramble J H Pasciak J E,Schatz A H.The construr tion of preconditioner for elliptic problrms by sub- structuring II, mathcmatic of computation, 1987,179 (49).1一16.
[13]BjØrstad P E,Widlund O B, iterative methods for the solution of elliptic problems on regions parti- tioned into substructures. SIAM Journal on Nu- merical Analysis. 1986,23(6):1097一1120.
[14]Hu X C,Chu D L. A preconditioner for elliptic prob- lems based on domain decomposition. Journal of Ts-
inghua University( Natural Science),1991,31(6):12 一19.(胡显承,储德林.求解椭圆型问题的一种基
于区域分解的预处理器.清华大学学报(自然科学),1991,31(6):12一19).
[15]Hu X C,Chu D L. Spectral analysis of the domain boundary operator contained in preconditioners for elliptic problems by domain decomposition. Journal of Tsinghua University(Natural Sci- ence),1992,32(3):8 -17.(胡显承,储德林.基于
区域分解的椭圆型问题预处理器中区域边界算子的谱分析.清华大学学报(自然科学),1992,32 (3) :8一17).
[16]Chu D L, Hu X C. A preconditioned conjugate gradient method witn nonoverlap domain decom- position. Journal of Computational Mathematics, 1993(1):58-68.储德林,胡显承.非重叠型区域分解预处理共辆梯度法.计算数学,1993(1); 58一68).
[17]Hu Q Y,Liang G P. A general framework to con- struct interface preconditioner for domain decom- position methods. Journal of Computational Mathematics, 1999 , 21(1) ; 117 -128.(胡齐芽,梁国平.区域分解界面预条件子构造的一般框架.
计算数学,1999,21(1):117一128).
No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!