Hi,

I am modeling the convective cooling flow inside an electronic enclosure (conjugated heat tranfer) using STAR-CCM+. I am trying to compare an intake with an exhaust fan, using velocity inlets with either possitive or negative inlet velocity. With the intake fan the convergence is quite ok, but when using the negative inlet velocity the turbulent quantities, turb. energy and disspation rate are not converging, also the other other flow quantities are converging. Does anyone know what to do about that? I varied the under-relaxation factors and also tried it with first order.

Thanks a lot!

The turbulence in CCM+ can be quite tricky to converge, especially in poly mehses. Check where the high residuals are by turning on temporary storage retention (on the solvers) and creating a threshold over the values. You may find that there is only one or two poor quality cells with high residuals and the rest are ok, also check convergence of the engineering data that you are interested in using some reports.

I am having a similar problem. On my simulation, everything converges except TDR and TKE. TKE and TDR residuals both flatten out a long ways above 1.0. Using the threshold,how high does the residual need to be to be considered too high?

[SB123]

Originally Posted by Josh
;172677

I am having a similar problem. On my simulation, everything converges except TDR and TKE. TKE and TDR residuals both flatten out a long ways above 1.0. Using the threshold,how high does the residual need to be to be considered too high?

I'm having similar problems with Tke and Sdr, created the threshold, don't know what value to use as a max.
they seem to dissipate to smaller value with each iteration though when it advances to the next timestep they jump back to the original(or close to it) value.
As to monitoring engineering values my value (drag) oscillates with each vortex shed as with the classic flow past a cylinder

[SB123]

Originally Posted by Josh
;172677

I am having a similar problem. On my simulation, everything converges except TDR and TKE. TKE and TDR residuals both flatten out a long ways above 1.0. Using the threshold,how high does the residual need to be to be considered too high?

i've since fixed my problem with high residuals. depending on your turbulence model - looks like K-eps? - you may need to adjust your turbulent boundary conditions for your inlet. K-eps is sometimes sensitive to inlet turbulence conditions depeding on which version you use ( i can't remember specifics on this). for this i would refer you to an article spalart wrote on the topic (tubulent boundary conditons)
you may also need to increase your inner iterations/decrease your timestep/ increase resolution for your mesh many different ways to do this. if you're using explicit unsteady, must monitor your CFL. if SS and you're still getting high resids and have looked elsewhere, try unsteady.

as for convergence, as many people have said, its not only dependant on residuals, physical values need to be monitored as well, but its also dependent on degree of accuracy you're looking for in your model. i would say generally people look for .001 convergence though depending on the problem, people can require residuals below 1e-8.

[irenefong92]

Originally Posted by SB123

i've since fixed my problem with high residuals. depending on your turbulence model - looks like K-eps? - you may need to adjust your turbulent boundary conditions for your inlet. K-eps is sometimes sensitive to inlet turbulence conditions depeding on which version you use ( i can't remember specifics on this). for this i would refer you to an article spalart wrote on the topic (tubulent boundary conditons)
you may also need to increase your inner iterations/decrease your timestep/ increase resolution for your mesh many different ways to do this. if you're using explicit unsteady, must monitor your CFL. if SS and you're still getting high resids and have looked elsewhere, try unsteady.

as for convergence, as many people have said, its not only dependant on residuals, physical values need to be monitored as well, but its also dependent on degree of accuracy you're looking for in your model. i would say generally people look for .001 convergence though depending on the problem, people can require residuals below 1e-8.

Hi. I need your advise. The residual graph just wont get lower even i have increase the inner iterations, and decrease the timesteps. My mesh is consider fine enough. I am using implicit unsteady. the inlet turbulent conditions looks okay. I dont know what i can do to improve the graph. So helpless.

Pls help. Thanks