【CFD理论】扩散项-02

非正交修正

最小修正（minimum correction）
正交修正（orthogonal correction）
超松弛 （over-relaxed）

非正交 non orthogonal

扩散项有限体积法回顾：
$$\begin{array}{rl}& {\int }_V(\nabla \cdot (D\nabla \varphi )+Q)dV=0\\ & \Rightarrow \sum _f{\int }_{f}D\nabla \varphi \cdot d\mathbf{S}+Q_p{V}_p=0\\ & \Rightarrow \sum _{f}D(\nabla \varphi {)}_f\cdot {\mathbf{S}}_f+Q_p{V}_P=0\end{array}$$

skewness和non-orthogonal的区别

网格正交的扩散项
$$(\nabla \varphi )\cdot \overrightarrow{S_f}=\frac{\left|\overrightarrow{S_f}\right|({\varphi }_N-{\varphi }_P)}{\left|\overrightarrow{d_{PN}}\right|}$$

三种分解方法：

1. 最小修正（minimum correction）
2. 正交修正（orthogonal correction）
3. 超松弛（over-relaxed）

minimum correction

$$\begin{array}{rl}& {\mathbf{S}}_f=\vec{\Delta }+\vec{k}\\ & \vec{\Delta }=\left|{\vec{S}}_f\right|\cdot cos\theta \cdot \frac{\overrightarrow{PN}}{\left|PN\right|}\\ & (\nabla \varphi {)}_f\cdot {\mathbf{S}}_f\\ & =(\nabla \varphi {)}_f\cdot \Delta +(\Delta \varphi {)}_f\cdot \mathbf{k}\\ & \end{array}$$

orthogonal correction approach

$$\begin{array}{rl}& {\mathbf{S}}_f=\vec{\Delta }+\vec{k}\\ & \vec{\Delta }=\left|{\vec{S}}_f\right|\cdot \frac{\overrightarrow{PN}}{\left|PN\right|}\\ & (\nabla \varphi {)}_f\cdot {\mathbf{S}}_f\\ & =(\nabla \varphi {)}_f\cdot \Delta +(\Delta \varphi {)}_f\cdot \mathbf{k}\\ & \end{array}$$

over-relaxed approach

$$\begin{array}{rl}& {\mathbf{S}}_f=\vec{\Delta }+\vec{k}\\ & \vec{\Delta }=\frac{\left|{\vec{S}}_f\right|}{cos\theta }\cdot \frac{\overrightarrow{PN}}{\left|PN\right|}=\frac{\overrightarrow{PN}{\left|{\mathbf{S}}_f\right|}^2}{{\mathbf{S}}_f\cdot \overrightarrow{PN}}\\ & (\nabla \varphi {)}_f\cdot {\mathbf{S}}_f\\ & =(\nabla \varphi {)}_f\cdot \Delta +(\Delta \varphi {)}_f\cdot \mathbf{k}\\ & \end{array}$$

Algebraic Equation for Non-orthogonal Meshes

$$\begin{array}{rl}& \sum _f(D\nabla \varphi )\cdot {\mathbf{S}}_f=\sum _f((D\nabla \varphi {)}_f\cdot (\Delta +\mathbf{k}))\\ & =\sum _f(D\nabla \varphi {)}_f\cdot \Delta +\sum _f(D\nabla \varphi {)}_f\cdot \mathbf{k}\\ & =\sum _f(D\frac{{\varphi }_N-{\varphi }_P}{PN})\left|\Delta \right|+\sum _f(D\nabla \varphi )\cdot \mathbf{k}\end{array}$$
$$\begin{array}{rl}& a_N=\frac{D}{PN}{\left|\Delta \right|}_f\\ & a_P=\sum _f\frac{D}{PN}{\left|\Delta \right|}_f\\ & b_P=S_P{V}_P+\sum _f(D\nabla \varphi )\cdot \mathbf{k}\\ & a_p{\varphi }_p+\sum _f{a}_N{\varphi }_N=b_P\end{array}$$

Example 2

1.
$$\begin{array}{rl}& \varphi =x^2+y^2+x^2{y}^2\\ & \nabla \varphi =\frac{\partial \varphi }{\partial x}\mathbf{i}+\frac{\partial \varphi }{\partial y}\mathbf{j}=(2x+2x{y}^2)\mathbf{i}+(2y+2x^{2}y)\mathbf{j}\\ & C(1.75,2)\\ & analyticalvalue:\\ & \nabla {\varphi }_C=(2\ast 1.75+2\ast 1.75\ast 2^2)\mathbf{i}+(2\ast 2+2\ast 2\ast 1.75^2)\mathbf{j}=17.5\mathbf{i}+16.25\mathbf{j}\\ & numericalvalue:\\ & \nabla {\varphi }_C=\frac{1}{V_C}\sum _f{\varphi }_f{\mathbf{S}}_f\\ & =\frac{1}{8.625}\left[84.625（2.5\mathbf{i}+\mathbf{j}）+42.0625（-\mathbf{i}+2\mathbf{j}）+4.3125（-2\mathbf{i}+0.5\mathbf{j}）+1.5（-\mathbf{i}+2\mathbf{j}）+12.37890625（1.5\mathbf{i}-1.5\mathbf{j}）\right]\\ & =20.63111\mathbf{i}+17.31454\mathbf{j}\end{array}$$

$$\begin{array}{rl}& \nabla {\varphi }_F=112.625\mathbf{i}+133.4375\mathbf{j}\end{array}$$

插值$\nabla {\varphi }_{f1}$:
$\nabla {\varphi }_{f1}=g_{f1}\nabla {\varphi }_F+(1-g_{f1})\nabla {\varphi }_C$
$\nabla {\varphi }_F=112.625\mathbf{i}+133.4375\mathbf{j}$
numerical value of the gradient at C:
$\nabla {\varphi }_C=20.63111\mathbf{i}+17.31454\mathbf{j}$
$g_{f1}=\frac{d_{Cf1}}{d_{Cf1}+d_{f1F}}=0.5$
numerical value:
$\nabla {\varphi }_{f1}=66.628055\mathbf{i}+75.37602\mathbf{j}$

analytical value:
$f_1(3,2.75)$
$\nabla {\varphi }_{f1}=(2x+2x{y}^2)\mathbf{i}+(2y+2y{x}^2)\mathbf{j}$
$\nabla {\varphi }_{f1}=51.375\mathbf{i}+55\mathbf{j}$

$-\nabla {\varphi }_{f1}\cdot {\mathbf{S}}_{f1}=E_{f1}\frac{{\varphi }_C-{\varphi }_F}{d_{CF}}-\nabla {\varphi }_{f1}\cdot {\mathbf{T}}_{f1}$
${\mathbf{S}}_{f1}=2.5\mathbf{i}+\mathbf{j}$
${\mathbf{e}}_1=\frac{{\mathbf{d}}_{CF}}{d_{CF}}=0.8575\mathbf{i}+0.51445\mathbf{j}$

(a) minimum correction approach
${\mathbf{E}}_{f1}=({\mathbf{e}}_1\cdot {\mathbf{S}}_{f1}){\mathbf{e}}_1=2.279\mathbf{i}+1.368\mathbf{j}\Rightarrow {\mathbf{E}}_{f1}=2.658$
${\mathbf{T}}_{f1}={\mathbf{S}}_{f1}-{\mathbf{E}}_{f1}=0.221\mathbf{i}-0.368\mathbf{j}\Rightarrow \nabla {\varphi }_{f1}\cdot {\mathbf{T}}_{f1}=(66.628055\mathbf{i}+75.37602\mathbf{j})\cdot (0.221\mathbf{i}-0.368\mathbf{j})=-13.014$
$-\nabla {\varphi }_{f1}\cdot {\mathbf{S}}_{f1}=\frac{2.658}{2.9155}({\varphi }_C-{\varphi }_F)+13.014=0.912({\varphi }_C-{\varphi }_F)+13.014$

(b) orthogonal correction approach
${\mathbf{E}}_{f1}={\mathbf{e}}_1\cdot {\mathbf{S}}_{f1}=2.309\mathbf{i}+1.385\mathbf{j}\Rightarrow {\mathbf{E}}_{f1}=2.693$
${\mathbf{T}}_{f1}={\mathbf{S}}_{f1}-{\mathbf{E}}_{f1}=0.191\mathbf{i}-0.385\mathbf{j}\Rightarrow \nabla {\varphi }_{f1}\cdot {\mathbf{T}}_{f1}=(66.628055\mathbf{i}+75.37602\mathbf{j})\cdot (0.191\mathbf{i}-0.385\mathbf{j})=-16.294$
$-\nabla {\varphi }_{f1}\cdot {\mathbf{S}}_{f1}=\frac{2.693}{2.9155}({\varphi }_C-{\varphi }_F)+16.294=0.924({\varphi }_C-{\varphi }_F)+16.294$

c) over-relaxed approach
${\mathbf{E}}_{f1}=\frac{{\mathbf{S}}_1\cdot {\mathbf{S}}_1}{{\mathbf{e}}_1\cdot {\mathbf{S}}_{f1}}{\mathbf{e}}_1=2.339\mathbf{i}+1.403\mathbf{j}\Rightarrow {\mathbf{E}}_{f1}=2.728$
${\mathbf{T}}_{f1}={\mathbf{S}}_{f1}-{\mathbf{E}}_{f1}=0.161\mathbf{i}-0.403\mathbf{j}\Rightarrow \nabla {\varphi }_{f1}\cdot {\mathbf{T}}_{f1}=(66.628055\mathbf{i}+75.37602\mathbf{j})\cdot (0.161\mathbf{i}-0.403\mathbf{j})=-19.649$
$-\nabla {\varphi }_{f1}\cdot {\mathbf{S}}_{f1}=\frac{2.728}{2.9155}({\varphi }_C-{\varphi }_F)+16.294=0.936({\varphi }_C-{\varphi }_F)+19.649$

bondary conditions for non-orthogonal grids

dirichlet boundary condition

neumann boundary condition

mixed boundary condition

skewness

${\varphi }_f={\varphi }_{f^{\prime }}+(\nabla \varphi {)}_{f^{\prime }}\cdot {\mathbf{d}}_{f^{\prime }f}$

under relaxation

$a_C{\varphi }_C+{\sum }_f{a}_N{\varphi }_N=b_P$
${\varphi }_p=\frac{b_P-{\sum }_f{a}_N{\varphi }_N}{a_P}$
显式：
${\varphi }_P={\varphi }_P^{old}+\alpha (\frac{b_p-{\sum }_f{a}_N{\varphi }_N}{a_P}-{\varphi }_P^{old})$
fields p 0.3
隐式：
$\frac{a_P}{\alpha }+{\sum }_f{a}_N{\varphi }_N=b_P+\frac{(1-\alpha )a_p}{\alpha }{\varphi }_p^{old}$
equations U 0.9

[CFD] Relaxation in CFD (Part 1) - Explicit Relaxation, Under-Relaxation Factor

【补充内容】Fluid Mechanics 101- non-orthogonal corrector2