数值求解Hilbert病态线性方程组.docx

病态线性代数方程组的求解理论的分析表明，求解病态的线性代数方程组是困难的。考虑方程组Hx=b的求解，其中H为Hilbert矩阵，，，1. 估计Hilbert矩阵2-条件数与阶数的关系；2. 选择问题的不同维数，分别用Gauss消去法，Jacobi迭代，GS迭代和SOR迭代求解，比较结果；3. 讨论病态问题求解的算法。解：1、取Hilbert矩阵阶数最高分别为n=20和n=100。采用Hilbert矩阵的2-条件数作为实验的比较对象，画出的曲线如下图所示：从图中可以看出，在n≤13之前，图像接近直线，在n＞13之后，图像趋于平缓，在一定的范围内上下波动。为了比较图像的线性部分，作出一条直线与已知曲线进行比较。比较直线的关系式为：，结果下图所示。从图2中可以看出，当n较小时，之间近似满足线性关系。当n继续增大到100时，关系图下图所示：从图中可以看出，图像的走势符合在n=20时的猜想，在n大于一定的值之后，图像趋于平缓，且在一定范围内震荡，同时又有一定上升趋势，但上升速度很慢。2、选择不同的阶数n，设方程组的精确解为xz=(1,1,…,1)T进行计算，用四种方法解x_Guass1、x_Jacobi1、x_GS1、x_SOR1对比表如下表所示。四种方法解x_Guass、x_Jacobi、x_GS、x_SOR对比表nx_Guassx_Jacobix_GSx_SOR31-8.22E+3070.9999911-1.552e+30810.999991-Inf0.999971411.0172e+30810.999991Inf0.999651.00011Inf1.00080.999741Inf0.999481.000251-6.07E+3070.998791.00011-1.2685e+3081.02230.999041-1.6592e+3080.905111.0041-Inf1.14210.994111-Inf0.930951.002861-6.64E+3070.9994911-1.4322e+3081.00590.999541-Inf0.995181.00111-Inf0.951491.00091-Inf1.10210.996211-Inf0.945361.002371-9.20E+3071.000711-Inf0.981921.00021-Inf1.10220.99791-Inf0.808241.00621-Inf1.0620.994641-Inf1.14480.998981-Inf0.899561.0021811.4378e+3081.00190.999971Inf0.958751.0011Inf1.19620.994081Inf0.730241.01141Inf0.954490.994641Inf1.18240.998061Inf1.14660.996681Inf0.828041.0042915.09E+3071.00310.9999211.1719e+3080.939461.001911.6249e+3081.24890.990291Inf0.754461.01540.99999Inf0.838240.995491Inf1.09850.998760.99998Inf1.22470.994051Inf1.11480.999691Inf0.775611.0046101-7.82E+3071.00360.999941-Inf0.93761.00111-Inf1.21420.995551-Inf0.889821.00450.99992-Inf0.779991.00111.0002-Inf0.958251.00030.99964-Inf1.16090.997481.0003-Inf1.22070.997760.99982-Inf1.08170.999271-Inf0.750191.0031117.97E+3071.00290.999951Inf0.958571.00070.99998Inf1.09130.998091.0002Inf1.09090.99980.99894Inf0.802911.0031.0037Inf0.828491.0010.99187Inf1.02360.999361.0111Inf1.18860.997760.99081Inf1.21280.997991.0042Inf1.06010.999610.99916Inf0.73672111E+3071.00120.999971Inf0.995921.00030.99993Inf0.919041.00021.0009Inf1.29740.996210.99316Inf0.887531.0041.0298Inf0.743091.00140.91765Inf0.