【CFD理论】对流项-05-非结构化网格

咔咔酷酷冲冲

已于 2022-08-29 10:51:58 修改

阅读量336

点赞数

分类专栏： of初阶文章标签：算法机器学习线性代数

于 2022-08-19 16:25:28 首次发布

本文链接：https://blog.csdn.net/zq93538196/article/details/126396968

版权

of初阶专栏收录该内容

16 篇文章 7 订阅

订阅专栏

【CFD理论】对流项-05

Unstructured Grids
discretisation in unstructured grids 非结构化网格离散

Unstructured Grids

discretisation in unstructured grids 非结构化网格离散

$\int_{CV}\frac{\partial}{\partial t}dV+\int_{CV}div(\rho \phi \boldsymbol u)dV=\int_{CV}div(\Gamma \ grad \ \phi)dV+\int_{CV}S_{\phi}dV$
在这里插入图片描述

$\boldsymbol {n,e_{\eta}}$ 分别是指向面外的法向和切向矢量
$\boldsymbol {PA}=\begin{matrix} x_A-x_P \\ y_A-y_P \end{matrix}\quad$
$\boldsymbol {ab}=\begin{matrix} x_b-x_a \\ y_b-y_a \end{matrix}\quad$
$\boldsymbol {e_\xi}=\frac{\boldsymbol {PA}}{|\boldsymbol {PA}|}$
$\boldsymbol {e_\eta}=\frac{\boldsymbol {ab}}{|\boldsymbol {ab}|}$
$\Delta \xi=|\boldsymbol {PA}|$
$\Delta \eta=|\boldsymbol {ab}|$
在这里插入图片描述
面i的面积 $\Delta A_i$
$\Delta A_i=\sqrt{\Delta x^2+\Delta y^2}$
面的单元法向量 $\boldsymbol n$
$\boldsymbol n=\begin{matrix} \frac{\Delta y}{\Delta A_i} , \frac{\Delta x}{\Delta A_i} \end{matrix}\quad$

discretisation of the diffusion term

扩散项离散，中心差分
$\int_{\Delta A_i}(\Gamma grad\phi)\cdot \boldsymbol n dA\approx\Gamma grad\phi\cdot\boldsymbol n_i\Delta A_i\approx\Gamma(\frac{\phi_A-\phi_P}{\Delta \xi})\Delta A_i$
之前的形式
$\int_f(\Gamma\Delta\phi)\cdot d\boldsymbol S=(\Gamma\Delta \phi)_f\cdot \boldsymbol n_fS_f=\Gamma(\frac{\phi_E-\phi_P}{\Delta x})_f S_f$
在这里插入图片描述

$\frac{\partial \phi}{\partial \xi}=\frac{\partial \phi}{\partial n}cos\theta-\frac{\partial \phi}{\partial \eta}sin\theta$
$\nabla \phi \cdot \boldsymbol n=\frac{\partial \phi}{\partial n}=\frac{\partial \phi}{\partial \xi}\frac{1}{cos\theta}+\frac{\partial \phi}{\partial \xi}tan\theta=grad\phi\cdot \boldsymbol n$
非结构化网格direct gradient 中心差分
$\frac{\partial \phi}{\partial \xi}=\frac{\phi_A-\phi_P}{\Delta \xi}$
非结构化网格cross-diffusion
$\frac{\partial \phi}{\partial \eta}=\frac{\phi_b-\phi_a}{\Delta \eta}$
$\nabla \phi \cdot \boldsymbol n\Delta A_i=\Delta A_i\frac{\phi_A-\phi_P}{\partial \xi}\frac{1}{cos\theta}+\frac{\partial \phi_b-\phi_a}{\partial \xi}tan\theta\Delta A_i$
$\frac{1}{cos\theta}=\frac{1}{\boldsymbol n\cdot\boldsymbol e_{xi}}=\frac{\boldsymbol n\cdot \boldsymbol n}{\boldsymbol n\cdot \boldsymbol e_{\xi}}$
$tan\theta=\frac{1}{\boldsymbol n\cdot\boldsymbol e_{xi}}=-\frac{\boldsymbol e_\xi \cdot \boldsymbol e_\eta}{\boldsymbol n\cdot \boldsymbol e_{\xi}}$
$\nabla \phi \cdot \boldsymbol n\Delta A_i=\Delta A_i\frac{\phi_A-\phi_P}{\partial \xi}\frac{\boldsymbol n\cdot \boldsymbol n}{\boldsymbol n\cdot \boldsymbol e_{\xi}}-\frac{\partial \phi_b-\phi_a}{\partial \eta}\frac{\boldsymbol e_\xi \cdot \boldsymbol e_\eta}{\boldsymbol n\cdot \boldsymbol e_{\xi}}\Delta A_i$
$\nabla \phi \cdot \boldsymbol n\Delta A_i=\Delta A_i\frac{\phi_A-\phi_P}{\partial \xi}\frac{\boldsymbol n\cdot \boldsymbol n}{\boldsymbol n\cdot \boldsymbol e_{\xi}}+S_{D-cross}$

Discretisation of the convective term

$\sum_f\int_f\rho\phi U\cdot d\boldsymbol S=\sum_f\int_{\Delta A_i}\rho\phi \boldsymbol u\cdot \boldsymbol n dA=\sum_f(\rho \phi\boldsymbol u)_f\cdot \boldsymbol n A_f$

linear upwind differencing

$\phi_e=\phi_P+\frac{\phi_P-\phi_W}{\Delta x}\frac{1}{2}\Delta x$
对于非结构化网格可以使用泰勒展开式将 $\phi$
$\phi(x,y)=\phi_P+(\nabla\phi)_P\cdot\Delta \boldsymbol r+\mathcal{O}(\Delta \boldsymbol r)^2$
如果是 $\Delta \boldsymbol r$ 是点P到面心的距离
$\phi_i=\phi_P+(\nabla \phi)_P \cdot \Delta \boldsymbol r$
$\Rightarrow$

对于非结构化网格，需要 $\color{red}面心值\phi_i$ ，就必须要知道 $\color{red}体心梯度(\nabla \phi)_P$
最小二乘法 $\color{red} 重构reconstruct$ 体心梯度

$\phi_1-\phi_0=(\frac{\partial \phi}{\partial x})|_0\Delta x_1+(\frac{\partial \phi}{\partial y})|_0\Delta y_1\\ \phi_2-\phi_0=(\frac{\partial \phi}{\partial x})|_0\Delta x_2+(\frac{\partial \phi}{\partial y})|_0\Delta y_2\\ \phi_3-\phi_0=(\frac{\partial \phi}{\partial x})|_0\Delta x_3+(\frac{\partial \phi}{\partial y})|_0\Delta y_3\\ \phi_4-\phi_0=(\frac{\partial \phi}{\partial x})|_0\Delta x_4+(\frac{\partial \phi}{\partial y})|_0\Delta y_4\\$
矩阵形式
$\begin{bmatrix} \Delta x_1&\Delta y_1\\ \Delta x_2&\Delta y_2\\ \Delta x_3&\Delta y_3\\ ...&...\\ \Delta x_N&\Delta y_N\\ \end{bmatrix} \begin{bmatrix} \Delta (\frac{\partial \phi}{\partial x})|_0\\ \Delta (\frac{\partial \phi}{\partial y})|_0\\ \end{bmatrix} = \begin{bmatrix} \phi_1-\phi_0\\ \phi_2-\phi_0\\ \phi_3-\phi_0\\ ...\\ \phi_N-\phi_0\\ \end{bmatrix}$
矩阵 $\color{red}超定 overdetermind$ ,所以不太可能直接求解
需要用 $\color{red}最小二乘法$
$\boldsymbol{A^TAx=A^Tb}\\ \boldsymbol{x=(A^TA)^{-1}A^Tb}$

TVD schmes in unstructured grids

结构化网格的TVD
$\phi_i=\phi_P+\frac{\psi(r)}{2}(\phi_E-\phi_P)\\ r=\frac{\phi_P-\phi_W}{\phi_E-\phi_P}$
$\color{red}在非结构化网格中，上游点W很难确定，所以需要构造一个假的点B$
在非结构化网格TVD

$r=\frac{2(\nabla \phi)_P\cdot \boldsymbol r_{PA}}{\phi_D-\phi_U}-1$
对流通量TVD表达式变为
$\phi_i=\phi_U=\frac{\psi(r)}{2}(\phi_D-\phi_U)\\ \psi(r)=0.25r+0.75(QUICK)\\ (\nabla \phi)_p最小二乘法求得\\ \phi_D下游，\phi_U上游$
在这里插入图片描述