一种求解大型Lyapunov矩阵方程的预处理并行算法

侯俊霞; 吕全义; 曹方颖; 谢公南

doi:10.3879/j.issn.1000-0887.2013.05.003

一种求解大型Lyapunov矩阵方程的预处理并行算法

doi: 10.3879/j.issn.1000-0887.2013.05.003

¹西北工业大学应用数学系,西安 710129;²西北工业大学机电学院,工程仿真与宇航计算技术联合实验室,西安 710072

基金项目: 国家自然科学基金资助项目（11202164）; 陕西省自然科学基金资助项目(2009JM1008)

详细信息

作者简介:
侯俊霞(1987—)，女，山东人，硕士生(E-mail:hjx2008cool@163.com)；吕全义(1963—)，女，沈阳人，副教授(通讯作者. E-mail:luquan@nwpu.edu.cn).

中图分类号: O246
计量
- 文章访问数: 1666
- HTML全文浏览量: 111
- PDF下载量: 1066
- 被引次数: 0
出版历程
- 收稿日期: 2013-03-22
- 修回日期: 2013-05-03
- 刊出日期: 2013-05-15

A Preconditioned Parallel Method for Solving Large Lyapunov Matrix Equation

¹Department of Applied Mathematics, Northwestern Polytechnical University,Xi’an 710129, P.R.China;²Engineering Simulation and Aerospace Computing, School of Mechanical Engineering, Northwestern Polytechnical University,Xi’an 710072, P.R.China

摘要

摘要: 研究了一种求解大型Lyapunov矩阵方程的并行预处理变形共轭梯度法．首先将处理小型矩阵方程的Smith预处理方法引入该问题的求解，将原矩阵方程转变为Stein方程，然后采用变形共轭梯度法并行求解预处理后的矩阵方程．其中遇到的难点是需要确定参数μ及求矩阵( A +μ I )的逆．基于估计特征值的Gerschgorin圆定理给出了参数μ的估值，再采用变形共轭梯度法并行求得矩阵( A +μ I )的逆，从而形成预处理后的矩阵方程．通过数值试验，该算法与未预处理的变形共轭梯度法相比较，预处理算法明显优于未预处理的算法，而且其并行效率高达0.85.
- Lyapunov矩阵方程 /
- 并行计算 /
- 变形共轭梯度法 /
- 预处理方法
Abstract: A parallel algorithm with preconditioned modified conjugate gradient method for solving large Lyapunov matrix equation. The preconditioned Simth method for small matrix equation was first introduced, and then the modified conjugate gradient method was used for parallel solving the preconditioned Stein matrix equation, which transformed from the original Lyapunov matrix equation. To fix the involved difficulties such as the determination of the parameter μ and the solving inverse matrix of the matrix ( A +μ I ),Gerschgorin theorem and the modified conjugate gradient method were employed. Several numerical experiments show the proposed algorithm is superior to the modified conjugate gradient without precondition. The parallel efficiency is up to 0.85.
- Lypunov matrix equation /
- parallel computation /
- modified conjugate gradient method /
- precondition method

HTML全文

参考文献(8)

[1]	徐树芳. 控制论中的矩阵计算[M]. 北京：高等教育出版社,2011.(XU Shu-fang.Matrix Computations in Control Theory [M].Beijing:Higher Education Press, 2011.(in Chinese))
[2]	李晓梅, 吴建平. 数值并行算法与软件[M]. 北京：科学出版社,2007.(LI Xiao-mei, WU Jian-ping.Numerical Parallel Algorithm and Software [M]. Beijing: Scince Press, 2007.(in Chinese))
[3]	Jbilou K.ADI preconditioned Krylov methods for large Lyapunov matrix equations[J].Linear Algebra and Its Applications, 2010,432(10): 2473-2485.
[4]	张宝琳,谷同祥,莫则尧. 数值并行计算原理与方法[M]. 北京：国防工业出版社,1999: 66-91.(ZHANG Bao-lin,GU Tong-xiang,MO Ze-yao.Principles and Methods of Numerical Parallel Computation [M]. Beijing: Scince Press,1999: 66-91.(in Chinese))
[5]	张凯院,徐仲. 数值代数[M]. 第二版修订版.北京：科学出版社,2010.(ZHANG Kai-yuan,XU Zhong.Numerical Algebra[M]. revised 2ed ed. Beijing: Scince Press,2010.(in Chinese))
[6]	陈国良,安虹,陈俊,郑启龙,单九龙. 并行算法实践[M]. 北京：高等教育出版社,2004.(CHEN Guo-liang,AN Hong,CHEN Jun,ZHENG Qi-long,SHAN Jiu-long.The Parallel Algorithm [M].Beijing: Higher Education Press,2004.(in Chinese))
[7]	Stykel T,Simoncini V. Krylov subspace methods for projected Lyapunov equations[J].Applied Numerical Mathematics,2012,62(1): 35-50.
[8]	曹方颖,吕全义,谢公南. 一类特殊矩阵方程的并行预处理并行共轭梯度算法[J]. 应用数学和力学,2013,34(3): 240-251.(CAO Fangying,Lü Quan-yi,XIE Gong-nan. A parallel preconditioned modified conjugate gradient method for a kind of matrix equation[J].Applied Mathematics and Mechanics,2013,34(3): 240-251.(in Chinese))