[akhshik]

Dear All

Have a nice day

Please find attached file, Results obtained after 715 iteration. As can be seen, the residual curve for some equation is above 1 (about 100). If the residual graph is higher than 1, what can be concluded? Does the model have a problem?

residual.JPG

[Deepak Bansal]

go to monitor option and go to normalization and instead auto u put as off..after that u check that its less than 1 or not.........if not less than 1 it means problem in ur set up.............

[Roman]

Actually, I don't think that this looks bad at all. It looks like your values converges well, what you should look out for are sudden "spikes" in your residuals that can push your solution in a divergence state.

Also, taken from manual:

"In general, use normalized residuals to monitor the behavior of the solvers at each iteration to make sure that the solution is converging toward the machine precision. If the simulation is steady state, the convergence is displayed over the total number of iterations. If the solution is time-dependent, make sure that convergence occurs within each time-step."

[flotus1]

If this is a transient simulation, decrease the time step size and increase the number of iterations per time step.

[akhshik]

Dear all,
thank you for your help

[kingmaker]

Hello all
I have been experiencing similar problem. The thing is .. I have sudden jumps in the residual especially for Tdr ... !

Also Tdr residual(normalization off) starts at 100 and decreases to 0.1 in 10 iterations and then every 40-50 iterations it jumps to 100 again and then starts decreasing for next 40-50 iterations and then again jumps to 100 ... this continues ... !! Rest all residuals are smooth and keep reducing ... in the order of 1e-5 or so .. ! only Tdr is oscillating above 1 or so .. !

I cannot figure out what is going on ... if there is mistake in my problem set up ... it would be very helpful if some one can give me some hints what could be wrong .

Thank you in advance
Aditya

[me3840]

Typically this occurs with an incorrect near-wall spacing or with incorrect inlet/initial turbulence values.

[hotash5]

you can try to reduce the Under relaxation factor (URF) of turbulence solver to 0.5 or 0.4.

[ashokac7]

Originally Posted by Deepak Bansal

go to monitor option and go to normalization and instead auto u put as off..after that u check that its less than 1 or not.........if not less than 1 it means problem in ur set up.............

Can someone explain why we need to deactivate normalization for tdr and tke residuals?

[bluebase]

Hi Ashok,


you don't have to deactivate it. It's just a different way to assess the residuals which might hide problems masked by a normalization, or fabricates false issues.



The normalization just takes the very first iteration's values to normalize all other residuals.
When you already start with a good solution, the first value is already very low, and residuals might won't drop well.


Quantities, such as tke, might have a low residual in the beginning. It then sometimes can happen that such a (relative) residual increases. However, the absolute (without normalization) might still be very low.

[ashokac7]

Originally Posted by bluebase

Hi Ashok,


you don't have to deactivate it. It's just a different way to assess the residuals which might hide problems masked by a normalization, or fabricates false issues.



The normalization just takes the very first iteration's values to normalize all other residuals.
When you already start with a good solution, the first value is already very low, and residuals might won't drop well.


Quantities, such as tke, might have a low residual in the beginning. It then sometimes can happen that such a (relative) residual increases. However, the absolute (without normalization) might still be very low.

Ok. Got it. Thank you very much Sebastian.