In 2D simulations using Large Eddy Simulation (LES) methodology, Favre averaging is commonly applied to the variables involved in the Navier-Stokes equations, resulting in: \begin{align}\label{aq} \frac{\partial (\overline{\rho} \tilde{u}_1)}{\partial t} + \frac{\partial (\overline{\rho} \tilde{u}_1\tilde{u}_1)}{\partial x_1} + \frac{\partial (\overline{\rho} \tilde{u}_1 \tilde{u}_2)}{\partial x_2} = - \frac{\partial \overline{p}}{\partial x_1} + \underbrace{\frac{\partial}{\partial x_1} \left( \overline{\tau}_{11} - \overline{\rho} \widetilde{u_1''u_1''} \right) + \frac{\partial}{\partial x_2} \left( \overline{\tau}_{12} - \overline{\rho} \widetilde{u_1''u_2''} \right)}_{(i)} \\ \frac{\partial (\overline{\rho} \tilde{u}_2)}{\partial t} + \frac{\partial (\overline{\rho} \tilde{u}_2\tilde{u}_1)}{\partial x_1} + \frac{\partial (\overline{\rho} \tilde{u}_2 \tilde{u}_2)}{\partial x_2} = - \frac{\partial \overline{p}}{\partial x_2} + \underbrace{\frac{\partial}{\partial x_1} \left( \overline{\tau}_{21} - \overline{\rho} \widetilde{u_2''u_1''} \right) + \frac{\partial}{\partial x_2} \left( \overline{\tau}_{22} - \overline{\rho} \widetilde{u_2''u_2''} \right)}_{(ii)}, \end{align} where $\tilde{u}_i$ and $u_i''$ respectively denote the mean and fluctuating components of the velocity $u_i$. Due to their complexity, the terms $(i)$ and $(ii)$ can be approximated by simpler expressions.

I came across a document that exclusively presents the non-dimensional form of the Navier-Stokes equations, employed for simulating compressible flow at low Mach numbers: \begin{align} \frac{\partial(\overline{\rho}^*\tilde{u}_1^*)}{\partial t^*} + \frac{\partial (\overline{\rho}^*\tilde{u}_1^* \tilde{u}_1^*)}{\partial x_1^*} + \frac{\partial (\overline{\rho}^*\tilde{u}_1^* \tilde{u}_2^*)}{\partial x_2^*} &= - \frac{1}{\text{Ma}^2} \frac{\partial \overline{p}^*}{\partial x_1^*} + \underbrace{\left(1+ \frac{\mu_t}{\nu} \right)\frac{1}{\text{Re}} \left(\frac{\partial^2 \tilde{u}_1^*}{\partial (x_1^*)^2} + \frac{\partial^2 \tilde{u}_1^*}{\partial (x_2^*)^2} \right)}_{(iii)} \\ \frac{\partial(\overline{\rho}^*\tilde{u}_2^*)}{\partial t^*} + \frac{\partial (\overline{\rho}^*\tilde{u}_2^* \tilde{u}_1^*)}{\partial x_1^*} + \frac{\partial (\overline{\rho}\tilde{u}_2^* \tilde{u}_2^*)}{\partial x_2^*} &= - \frac{1}{\text{Ma}^2} \frac{\partial \overline{p}^*}{\partial x_2^*} + \underbrace{\left(1+ \frac{\mu_t}{\nu} \right)\frac{1}{\text{Re}} \left(\frac{\partial^2 \tilde{u}_2^*}{\partial (x_1^*)^2} + \frac{\partial^2 \tilde{u}_2^*}{\partial (x_2^*)^2} \right)}_{(iv)}, \end{align} where $\mu_t$ is the turbulent viscosity and $\nu$ the kinematic viscosity. I know that the nondimensional equations are obtained from the primitive equations through a substitution of variables: \begin{equation} x_1^* = \frac{x_1}{L}, \, \, x_2^{*} = \frac{x_2}{L}, \, \, t^{*} = \frac{U t}{L}; \,\, \tilde{u}_1^{*} = \frac{\tilde{u}_1}{U}, \, \, \tilde{u}_2^{*} = \frac{\tilde{u}_2}{U}; \, \, \bar{\rho}^{*} = \frac{\bar{\rho}}{{\rho}_{\infty}}, \, \, \bar{p}^{*} =\frac{\bar{p}}{p_{\infty}}, \, \, \text{Re} = \frac{ \rho_{\infty} U L}{\mu_{\infty}} = \frac{UL}{\nu}; \,\, \text{Ma} = \frac{U}{\sqrt{p_{\infty}/\rho_{\infty}}}. \end{equation} Is there a way to deduce the expression used to approximate the diffusive terms $(i)$ and $(ii)$ of the equations (in its primitive form), so that after the non-dimensionalization procedure these terms become respectively $(iii)$ and $(iv)$?


1 Answer 1


Focusing only on diffusion of the average flow and turbulent terms,

\begin{equation} \nabla \cdot \left[ \mathbb{S} + \rho \widetilde{\mathbf{u}' \mathbf{u}'} \right] \ . \end{equation} Writing

  • the average viscous stress tensor in terms of the deformation velocity, \begin{equation} \mathbb{S} = 2 \mu \mathbb{D} = 2 \mu \frac{1}{2}(\nabla \mathbf{u} + \nabla^T \mathbf{u}) = \mu \left(\nabla \mathbf{u} + \nabla^T \mathbf{u} \right) = \dfrac{\rho \nu U}{L} \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) \end{equation}

  • the turbulent stress with a turbulent viscosity model \begin{equation} \rho \widetilde{\mathbf{u}' \mathbf{u}'} = 2 \mu_t \mathbb{D} = \dots = \dfrac{\rho \nu_t U}{L} \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) = \dfrac{\nu_t}{\nu} \dfrac{\rho \nu U}{L} \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) \ . \end{equation}

Thus, performing the non-dimensionalization in the usual way and dividing by $\frac{\rho U}{T} = \frac{\rho U^2}{L}$ coming from the convective term, you get \begin{equation}\begin{aligned} \dots & = \dfrac{L}{\rho U^2} \dfrac{1}{L} \nabla^* \cdot \left[ \dfrac{\rho \nu U}{L} \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) + \dfrac{\nu_t}{\nu} \dfrac{\rho \nu U}{L} \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) \right] = \\ & = \dfrac{\nu}{U L} \left( 1 + \dfrac{\nu_t}{\nu} \right) \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) = \\ & = \dfrac{1}{\text{Re}} \left( 1 + \dfrac{\nu_t}{\nu} \right) \left(\nabla^* \mathbf{u}^* + \nabla^{*T} \mathbf{u}^* \right) \end{aligned}\end{equation}


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.