【数学建模笔记 20】数学建模的偏微分方程的数值解

20. 偏微分方程的数值解

定解问题

各种物理性质的定常过程都可用椭圆型方程描述
$$\Delta u=\frac{{\partial }^{2}u}{\partial x^2}+\frac{{\partial }^{2}u}{\partial y^2}=f(x,y),$$
当 $f(x,y)=0$ 时，即拉普拉斯方程
$$\Delta u=\frac{{\partial }^{2}u}{\partial x^2}+\frac{{\partial }^{2}u}{\partial y^2}=0.$$
第一边值问题
$$\{\begin{array}{rl}& \frac{{\partial }^{2}u}{\partial x^2}+\frac{{\partial }^{2}u}{\partial y^2}=f(x,y),(x,y)\in \Omega ,\\ & u(x,y){|}_{(x,y)\in \Gamma }=\phi (x,y),\Gamma =\partial \Omega .\end{array}$$
其中 $\Omega$​ 为以 $\Gamma$ 为边界的有界区域，$\Gamma$ 为分段光滑曲线，$\Omega \cup \Gamma$ 称定解区域，$f(x,y),\phi (x,y)$ 分别为 $\Omega ,\Gamma$ 上的已知连续函数。

差分解法

椭圆型

考虑第一边值问题
$$\{\begin{array}{rl}& \frac{{\partial }^{2}u}{\partial x^2}+\frac{{\partial }^{2}u}{\partial y^2}=f(x,y),(x,y)\in \Omega ,\\ & u(x,y){|}_{(x,y)\in \Gamma }=\phi (x,y),\Gamma =\partial \Omega .\end{array}$$
取 $h,\tau$ 分别为 $x,y$​ 方向的步长，记
$$(k,j)=(x_k,y_j),u(k,j)=u(x_k,y_j),f_{k,j}=f(x_k,y_j),$$
则由二阶差商公式
$$\frac{{\partial }^{2}u}{\partial x^2}{|}_{k,j}=\frac{u(k+1,j)-2u(k,j)+u(k-1,j)}{h^2}+O(h^2),$$

$$\frac{{\partial }^{2}u}{\partial y^2}{|}_{k,j}=\frac{u(k,j+1)-2u(k,j)+u(k,j-1)}{{\tau }^2}+O({\tau }^2)$$

于是有
$$\frac{u(k+1,j)-2u(k,j)+u(k-1,j)}{h^2}$$

$$+\frac{u(k,j+1)-2u(k,j)+u(k,j-1)}{{\tau }^2}$$

$$=f_{k,j}+O(h^2+{\tau }^2).$$

略去 $O(h^2+{\tau }^2)$ 得近似的差分方程
$$\frac{u_{k+1,j}-2u_{k,j}+u_{k-1,j}}{h^2}+\frac{u_{k,j+1}-2u_{k,j}+u_{k,j-1}}{{\tau }^2}=f_{k,j}.$$
称五点菱形格式。

实际计算时常取 $h=\tau$，得
$$\frac{1}{h^2}(u_{k+1,j}+u_{k-1,j}+u_{k,j+1}+u_{k,j-1}-4u_{k,j})=f_{k,j},$$
记
$$\frac{1}{h^2}♢u_{k,j}=f_{k,j},$$
其中
$$♢u_{k,j}=u_{k+1,j}+u_{k-1,j}+u_{k,j+1}+u_{k,j-1}-4u_{k,j}.$$

双曲型

考虑
$$\frac{{\partial }^{2}u}{\partial t^2}=a^2\frac{{\partial }^{2}u}{\partial x^2}.$$
令 $v_1=\frac{\partial u}{\partial t},v_2=\frac{\partial u}{\partial x}$​，则有
$$\{\begin{array}{rl}& \frac{\partial v_1}{\partial t}=a^2\frac{\partial v_2}{\partial x},\\ & \frac{\partial v_2}{\partial t}=a^2\frac{\partial v_1}{\partial x}.\end{array}$$
记 $v=(v_1,v_2{)}^T$，则有
$$\frac{\partial v}{\partial t}=\left(\begin{array}{cc}0& a^2\\ 1& 0\end{array}\right)\frac{\partial v}{\partial x}=A\frac{\partial v}{\partial x}.$$
存在矩阵 $P$，使得
$$PA{P}^{-1}=\left(\begin{array}{cc}a& 0\\ 0& -a\end{array}\right)=\Lambda .$$
作变换 $w=Pv=(w_1,w_2{)}^T$，则有
$$\frac{\partial w}{\partial t}=\Lambda \frac{\partial w}{\partial x}.$$
而对于一阶双曲型方程，取 $x,t$ 方向步长分别为 $h,\tau$，则有
$$\frac{u_{k,j+1}-u_{k,j}}{\tau }+a\frac{u_{k+1,j}-u_{k,j}}{h}=0,$$

$$\frac{u_{k,j+1}-u_{k,j}}{\tau }+a\frac{u_{k,j}-u_{k-1,j}}{h}=0,$$

$$\frac{u_{k,j+1}-u_{k,j}}{\tau }+a\frac{u_{k+1,j}-u_{k-1,j}}{2h}=0.$$