Strong oscillation at transonic mach number
I'm conducting a study with fluent to test drag decomposition method. The goal is to be able to tell if Fluent his able to give good result for drag using other techniques that the so called one Nearfield. I'm conducting my study on the well known NACA0012 with zero angle of attack using a regular C-H type mesh with mach number variating beetween 0.72 to 0.85. As experiment predict, the shock appear around mach 0.76 and become stronger and stronger has the mach number increase. For mach number up to 0.82, I was able to obtain a converged solution up to 1e-14. The results for the drag that I obtained showed a well agreement with litterature.
But here's my problem, when I try to converge my solution for mach number of 0.85, I'm getting strong oscillation at 1x10e-2 (residuals). I'm using the Roe scheme with second order upwind discretisation with a courant number of 1 (in fact, I tried courant number beetween 1 to 5 with the same problem). I refined the mesh by double it both x and y direction and I'm getting the same oscillations around 1x10e-5. I doubled the mesh again and obtained the same oscillation a 1x10-7. So I changed my techniques (cause refined the mesh again will be prohibitive in computational time) and tried to refined the mesh with entropy gradient (so the mesh was refined at shock location and at the leading edge) Again the same problem. Does any one have an idea how to get rip off this problem?
By the way, I forgot to tell that I'm working with Euler inviscid equations.
Thanks for your support.
Thanks for you're link but I already read it. I'm afraid that there is a lot of things that is not true in this topics and anyway it's more about viscous flow. Mine is inviscid. The other thinks his I'm pretty sure my grid his fine cause I ran a lot of simulation with it for mach number below or equal to 0.82 without encountering any problem. I should then reformulate my question:
Knowing the behaviour of Fluent, does anybody know a way to make it converge when there is a strong shock ? (maybe converge at first order till a certain accuracy then swith to second order, using an higher order discretisation scheme like Quick, etc...)
Another question, wich kind of second order upwind does Fluent use?? Is there kind of TVD with high resolution scheme or it's simple second order upwind?
In advance, thanks for your support.
|All times are GMT -4. The time now is 22:06.|