[miguel.salazar]

I am modeling the microfluidic device in the following picture:

Screenshot 2023-03-03 at 09.04.37.jpg
It is comprised of an inlet vertical manifold with a cross section of 1cm by 1mm, a horizontal network of microchannels with cross section of 100 micrometers width and 300 micrometers height and an outlet with same dimensions as the inlet.

I am imposing a pressure difference between inlet and outlet mouth (red arrows) of 200 mbar, which I predict it will result in a flow rate of 20 ml/s more or less. The Reynolds number is only of a couple of hundreds. The main issue however is the 90 degrees corner at the bottom of the outlet. The flow exists the microchannels at high velocity (more than 1 meters per second) and this results in a complex series of vortices at the bottom of the outlet. I have learned that this situation is similar to a jet impingement. Unfortunately, I cannot modify the geometry to facilitate the flow. It is a sharp 90 degrees indeed.

I have been researching turbulence models to solve this issue. What I have learned is that none of the eddy viscosity models is appropriate for this scenario. What is the right method for this jet impingement? An important constraint that I have is that I need the method to be stationary because it will be embedded in an optimization routine and I cannot afford to resolve transient modes. My understanding is that LES models are not necessarily steady, which would leave them out.

[FMDenaro]

Originally Posted by miguel.salazar

I am modeling the microfluidic device in the following picture:

Attachment 93582
It is comprised of an inlet vertical manifold with a cross section of 1cm by 1mm, a horizontal network of microchannels with cross section of 100 micrometers width and 300 micrometers height and an outlet with same dimensions as the inlet.

I am imposing a pressure difference between inlet and outlet mouth (red arrows) of 200 mbar, which I predict it will result in a flow rate of 20 ml/s more or less. The Reynolds number is only of a couple of hundreds. The main issue however is the 90 degrees corner at the bottom of the outlet. The flow exists the microchannels at high velocity (more than 1 meters per second) and this results in a complex series of vortices at the bottom of the outlet. I have learned that this situation is similar to a jet impingement. Unfortunately, I cannot modify the geometry to facilitate the flow. It is a sharp 90 degrees indeed.

I have been researching turbulence models to solve this issue. What I have learned is that none of the eddy viscosity models is appropriate for this scenario. What is the right method for this jet impingement? An important constraint that I have is that I need the method to be stationary because it will be embedded in an optimization routine and I cannot afford to resolve transient modes. My understanding is that LES models are not necessarily steady, which would leave them out.

First of all, why do you think about turbulence when you wrote near laminar? The Reynolds number is quite small, you can suppose the flow to be laminar (eventually unsteady).



Second, the flow is unsteady in its nature? LES (and DNS) can be performed for that condition but you must solve the 3D and unsteady flow problem.


If you are not able to do that, your only chance is to use a statistical steady RANS formulation. Depending on the geometry and BC you could think to use a 2D assumption. However, RANS would badly work for a laminar flow and you have to renounce to any description of fine vortical structures. But a rude solution could work for your optimization problem.

[miguel.salazar]

Originally Posted by FMDenaro

First of all, why do you think about turbulence when you wrote near laminar? The Reynolds number is quite small, you can suppose the flow to be laminar (eventually unsteady).



Second, the flow is unsteady in its nature? LES (and DNS) can be performed for that condition but you must solve the 3D and unsteady flow problem.


If you are not able to do that, your only chance is to use a statistical steady RANS formulation. Depending on the geometry and BC you could think to use a 2D assumption. However, RANS would badly work for a laminar flow and you have to renounce to any description of fine vortical structures. But a rude solution could work for your optimization problem.

Thanks for your response. I basically would like to have a way to model the flow such that I do not need a highly refined mesh in the 90 degrees corner and I can calculate a steady-state solution. I would like this model to capture the correct global flow-rate and pressure drop. I understand now that this scenario is not necessarily turbulence because it might not be chaotic, but at the same time, it is computationally demanding due to the fine vortices that I would like not accurately resolve.

[Johncrunk11]

Simulation of laminar regions with LES method may provoke the problem of unphysical fluctuations. In order to create turbulence at the inlet, the velocity field of the DNS of a channel flow for which the bulk velocity was 0.6 m/s, is scaled and implemented as inlet boundary condition for velocities. No slip boundary condition is used for all solid boundaries.
For the temperature, Dirichlet boundary condition is applied at the hot tube and the inlet. The temperature is set to 80°C and 25°C on the hot tube and at the inlet respectively. Except for the outlet, homogeneous Neumann boundary condition is used over all remaining boundaries.
At the outlet, for both temperature and velocities, a convective boundary condition (Sohankar et al., 1998) is applied.
The dynamic viscosity and the Prandtl number of the fluid are μ = 18.9 × 10− 6 and Pr = 0.7, respectively.
In order to accelerate the convergence of the numerical computations to a fully developed condition, either the results of a 2D-RANS simulation or whenever existed, the results of previous simulations for different grid configuration, were interpolated and applied as initial boundary condition