[juste]

hi guys

i am trying to simulate an open channel flow (like simulating a river in a lab flume) in fluent..but without using vof...instead i am defining the top surface as a moving wall....so that the air-water interface is preserved to some extent...i am not interested in the interface and hence not using vof as of now...

the problem i am facing is that the residuals are not converging properly....initially they start converging, then start diverging, then flat, and then all goes weird.....please have a look at the attached pic...

however, the velocity profile and turbulence KE obtained from this simulation is very close to the experimental data i obtained from the lab flume.... i am just simulating this to verify and validate my model first...

do you have any idea why this is happening...any suggestions/solutions...

mesh details:
i guess i am already having good enough mesh
1. it is structured hexahedral mesh
2. aspect ratio (AR) = 1 - 178 (higher AR in boundary layer only, used inflation)
3. orthogonality = 0.16 -1.0 (so no skewed cells)
4. total cells = 1.4 million
5. i have used inflation for the boundary layer

fluent details:
in fluent also, i guess i am using pretty simple stuff
1. pressure based, steady state, gravity enabled solver
2. turbulence model - kepsilon with scalable wall function
3. material = liquid-water

4. inlet BC = mass-inflow, specified turbulence intensity and hydraulic dia
5. otlet BC = outflow
6. bottom and side = wall no slip
7. top = moving wall as mentioned before

8. urf - default values
9. SIMPLEC / least square / all second order discretization methods used

i am not sure why the residuals are behaving like this, although the output velocity and turbulence are matching...the geometry as well as the model is pretty simple (just a rectangular flume with a cavity) so i guess there shouldn't be much of a problem..

please let me know if you all have any ideas...thanks a lot..

[edoan]

try starting with first order discretization then switching up to second order after 500 or so iterations

[juste]

hi ethan and desmond, thanks for both of your replies

i just changed the mesh and now it is converging very nicely, 1e-6 residuals in 2700 iterations...i guess that's okay with me..

i didn't change any other thing in fluent...but i wonder why was it behaving so weird previously...

earlier too i had a hex mesh (using cutcell and inflation layers) and now I am having hex mesh too but using edge sizing and biased divisions, no inflation...

anyway, it has converged now....

thanks so much for both of your replies.

[bluebird786]

Hi gentlemen,

I am new to this subject and this forum so please take it easy on my ignorance.

I am trying to run a few supersonic cases on a jet model. Due to the time constraint I didnt want to create a very fine mesh. These are the current residuals graph.

[IMG][/IMG]

I know it doesnt show converging results. Is the only solution for me is refining mesh. If i do it causes the iterations to become really slow and I dont have a week to run this. Any other options for a quick conversion. I mean i can run the simulation for a day at most. Any help is highly appreciated. Please feel free to give any comments.