Elliptic Blending Model


The model implemented here is the one described in Thielen et al. (Int Journal of Heat and Mass Transfer, 48, 2005, pp.1583-1598).

The model solves the full set Reynolds stresses %$\overline{u_i u_j}$%, a dissipation equation %$\varepsilon$% and an elliptic equation for %$\alpha$%.

The Reynolds stress equation can be written as:

%BEGINLATEX% \begin{align} \frac{D\overline{u_i u_j}}{Dt} = &P_{ij} + {\Phi^*}_{ij} - \varepsilon_{ij}+\frac{\partial}{\partial x_k}\left[ (\nu \delta_{kl} + C_s\overline{u_ku_l}\tau) \frac{\partial \overline{u_i u_j}}{\partial x_l} \right] \\ \frac{D\varepsilon}{Dt} = & \frac{C'_{\varepsilon 1}P - C_{\varepsilon 2} \varepsilon}{\tau}+\left[ (\nu \delta_{kl} + C_{\varepsilon}\overline{u_ku_l}\tau) \frac{\partial \varepsilon }{\partial x_l} \right] \end{align} %ENDLATEX%

The main idea of the model is to blend the near wall part with the farfiled one via %$\alpha$%, which is defined as:

%BEGINLATEX% \begin{equation} \alpha-L^2\nabla^2\alpha = 1 \end{equation} %ENDLATEX%

With this parameter, the equation for the velocity-pressure-gradient correlation %$\Phi^*_{ij}$% can be written as:

%BEGINLATEX% \begin{equation} \Phi^*_{ij} = (1-\alpha^2)\Phi^w_{ij}+\alpha^2\Phi^h_{ij} \end{equation} %ENDLATEX%

The "homogeneous" part is taken from the SSG (Speziale et al.) %BEGINLATEX% \begin{align*} \Phi^h_{ij} =& -\left( C_1 +C_2\frac{P}{\varepsilon}\right) \varepsilon a_{ij} + C_3kS_{ij}\&+C_4k \left(a_{ik}S_{jk} +a_{jk}S_{ik} -\frac{2}{3}\delta_{ij}a_{lk}S_{kl} \right)\&+C_5k \left( a_{ik}\Omega_{jk} + a_{jk}\Omega_{ik} \right) \end{align*} %ENDLATEX%

And the near wall part is derived to satisfy the near wall asymptotic behaviour and stress budgets:

%BEGINLATEX% \begin{equation*} \Phi^w_{ij} = -5\frac{\varepsilon}{k} \left( \overline{u_i u_k} n_j n_k + \overline{u_j u_k} n_i n_k - \frac{1}{2} \overline{u_k u_l}n_k n_l (n_i n_j + \delta_{ij} ) \right) \end{equation*} %ENDLATEX%

with the following definitions:

%BEGINLATEX% \begin{align*} &a_{ij} = \frac{\overline{u_i u_j}}{k}-\frac{2}{3}\delta_{ij}; \quad S_{ij} = \frac{1}{2}\left(\frac{\partial U_i}{\partial x_j} + \frac{\partial U_j}{\partial x_i} \right); \quad \Omega_{ij} = \frac{1}{2}\left(\frac{\partial U_i}{\partial x_j} - \frac{\partial U_j}{\partial x_i} \right) \\ &\mathbf{n} = \frac{\nabla \alpha}{\parallel\nabla \alpha\parallel}; \quad \tau = \max \left( \frac{k}{\varepsilon}; \, C_{\tau} \sqrt[2]{\frac{\nu}{\varepsilon}} \right); \quad L=C_L\max \left( \frac{k^{3/2}}{\varepsilon}; \, C_{\eta} \sqrt[4]{\frac{\nu^3}{\varepsilon}} \right) \end{align*} %ENDLATEX%

The constants used are:

%BEGINLATEX% \begin{tabular}{|c c c c c c c c c c c|} \hline $C_s$ & $C_{\varepsilon}$ & $C_{\varepsilon}^0$ & $C_{\varepsilon2}$ & $C_1$ & $C_2$ & $C_4$ & $C_5$ & $C_{\tau}$ & $C_L$ & $C_{\eta}$ \\ \hline 0.21&0.18&1.44&1.83&1.7&0.9&0.625&0.2&6.0&0.161&80 \\ \hline \end{tabular} \\ \centering $C_{\varepsilon 1} = C_{\varepsilon 1}^0 \left(1+0.03(1-\alpha^2)\sqrt{\frac{k}{\overline{u_i u_j} n_i n_j}}\right)$ %ENDLATEX%

-- JuanUribe - 2009-11-05

Current Tags:
create new tag
, view all tags
Topic attachments
I Attachment Action Size Date Who Comment
elsegz ebm_channel.tar.gz manage 283.5 K 2009-11-05 - 12:41 JuanUribe File for the EBM for a channel flow calculation (v1.4.0)
Topic revision: r3 - 2009-11-06 - 10:17:22 - FlavienBillard
Saturne Web
19 Sep 2020
Test Cases

tip Saturne Tip of the Day
Channel flow statistics
Channel flow statistics have been developed and implemented for Code Saturne Version 1.3.1 RichardHoward ... Read on Read more


Code_Saturne collaborative website
@ the University of Manchester
Copyright & by the contributing authors. Unless noted otherwise, all material on this web site is the property of the contributing authors.