Looking for a solver (Mach 0.4, Turbulent, Heat Transfer, Second Order)
I'm looking for some advice to help me to choose an OF2 solver. I need to run some air flow simulations (Cylinder, Sphere, Naca Airfoil, ...) with a mach number below 0.5 (mostly Mach 0.2 to 0.35).
The important thing is that this solver must include:
- a turbulence model (k-e / k-w / ...)
- a good thermo model to compute the heat flux
- a second order scheme
Other things (but I can change it myself), it has to include:
- a dynamic mesh motion
My biggest concern is about the steadiness of the solver.
For the cases including a cylinder or a sphere, the flow is often unsteady... However, keeping a Courant number below 0.5 with a fine mesh and a high speed often lead to very small time step. The problem is that I will couple this solver with another home made solver to make simulation of a couple of minutes (running 5 minutes with time step of 1e-6 is not really acceptable). So I tend to focus on a steady solver even my flow can be in some cases unsteady (my simulation will become a succession of steady states).
And to be the perfect solver, it has to be robust with an "easy" convergence process.
What do you think? Should I try the Compressible solvers (rhoSimpleFoam, rhoPimpleFoam, sonicDyMFoam, rhoCentralFoam) or the Heat transfer solvers (buoyantSimpleFoam, buoyantBoussinesqSimpleFoam) or just an Incompressible solver and add a thermal equation (simpleFoam) ?
Do you think that at Mach 0.4 the compressibility is already important for the Heat Flux over a Naca Airfoil ?
I'm not an expert in compressible flows, but I'm almost certain that compressibility is not negligible at Mach 0.4. So if I were you, I would certainly start from a compressible solver.
Regarding the steadiness of the solver, I would recommend a pimple-based solver. Simply because it behaves stable at larger courant numbers than the piso-solvers (you should easility be able to go up to Co = 2...5). Your time step will benefit from it ;-). Something that also can help to increase the convergence speed is to use the localEuler ddt-scheme. This will optimize the time step for each cell which is allowed since your are applying a RANS model to obtain a steady state.
Hope that this already helps you a bit
So following your advice I should focus on "rhoPimpleFoam" using the Pimple algorithm coupled with a localEuler time scheme.
Just a stupid question, can a localEuler scheme be used in the case of a "dynamic" simulation. For example, what happen to a cylinder having von karman vortex ? Did you use this scheme before ?
Edit, I'm asking this because most of the topics I've seen talk about reaching a steady state and don't talk about the dynamic of the flow.
The localEuler time scheme sets the time step in each cell separately based on the local Courant number (if I'm not mistaken). This basically means that the time evolution for each cell is different every calculation step you make. So you should not use this time stepping procedure for dynamic simulations.
I never used it myself so I can't explain how to do it exactly, but this might help you:
|All times are GMT -4. The time now is 10:07.|