[sarrazs]

Hi,
I'm doing a transient DNS analysis for an isothermal bubble column. Total time = 150s and the time step = 0.01s. In the output control for Trn results when I change the output frequency form every time step to Time interval=10s. The flow configuration is very different in the two cases. Is there any logical explanation?? Should I select the every time step option to have more accurate results?
Thank you

[ghorrocks]

0.01s does not sound like DNS time scales. Where did you establish that time step from?

Writing results files does not affect the simulation. If the simulation result has changed then you changed something in the setup.

[sarrazs]

thank you for your reply.
For the time step, it was big just to do a test. I did the test an other time and I found out that the trns result frequency doesn't affect the final result. I'm figuring out what did I change.

[ghorrocks]

By DNS you mean direct numerical simulation, don't you? Have you confirmed your mesh and numerics can resolve the Kolmogorov length and time scales? This is very challenging in CFX, it is not a solver which is designed for DNS simulations.

Or do you mean something else by DNS?

[urosgrivc]

output control Trn results can and does not influence your results.

that is just to limit the size of the output file if you dont need results at every timestep, you will have only 15 timesteps in cfx post if You chose 10s interval for your 150s long run and if you chose every timestep than you will have 150000 result timesteps and your result file will be aproximately 1000 times larger. and for true dns this will be yet another few fthousant times larger.

Are you talking about 150s long DNS run? you do know that quantum computers are not here jet .
Oh anyway I think thats imposible.
I wonder what the world record is for the logest dns run?

If you are realy doing a DNS than the problem with your results is you dont have none and any litle change to the BCs, mesh or domain changes your results as your simulation is not independant of all the wariables (trn result frequency does not).
DNS is worlds most expensive way of doing cfd and is not usefull for the normal user.
You will probably need to use some kind of turbulence model.

nice reading_dns;
https://en.wikipedia.org/wiki/Direct...cal_simulation
or maybe LES
https://en.wikipedia.org/wiki/Large_eddy_simulation