[hotboy]

Hello friends!
Why it's difficult to converged when we use the fluent to simulation,could you recomend some paper,books or idea about this to me ?Thank you very much!

[LuckyTran]

Originally Posted by hotboy

Hello friends!
Why it's difficult to converged when we use the fluent to simulation,could you recomend some paper,books or idea about this to me ?Thank you very much!

It's difficult because it's highly non-linear: buoyancy/density depends on temperature, temperature depends on natural convection, natural convection depends on.... density again. Using boussinesq model may help as well a having a very good initial guess. Reducing under-relaxation factors on density and temperature may help stabilize it if you are having divergence problems.

[hotboy]

Originally Posted by LuckyTran

It's difficult because it's highly non-linear: buoyancy/density depends on temperature, temperature depends on natural convection, natural convection depends on.... density again. Using boussinesq model may help as well a having a very good initial guess. Reducing under-relaxation factors on density and temperature may help stabilize it if you are having divergence problems.

Dear LuckyTran!Thank you very much !The temperature differences in my domain are large,so I can't use boussinesq model .And I read the ”ansys help“which prompts that "No steady-state solution exists for some laminar,high-Rayleigh-mumber flows "If my case is big,it will take a lot of time to run ,what can I do?

[Bionico]

Hello hotboy,
I worked on natural convection in closed domain: I had the same difficulties as yours and I found useful some solver set up like Density-based solver (Iterations with Solution Steering) or Pressure-based-coupled with pseudo-transient. They are heavier in terms of memory, but they even require less iterations to converge than segregated solvers!

Best Regards

[hotboy]

Originally Posted by Bionico

Hello hotboy,
I worked on natural convection in closed domain: I had the same difficulties as yours and I found useful some solver set up like Density-based solver (Iterations with Solution Steering) or Pressure-based-coupled with pseudo-transient. They are heavier in terms of memory, but they even require less iterations to converge than segregated solvers!

Best Regards

Hi Bionico!Thank you very much!If your suggestion can't work I have no choice but to use the unsteady approch?I don't want to use the unsteady approch,because my case is big that it will take a lot of time !

[Bionico]

Unfortunately transient simulation would be the best way because the system is intrinsically unsteady. I was able to use those solvers with steady simulation, they took some time to tune up the numerical parameters (CFL, etc...), but they worked well.
If you will have any question about the specific solver I may help you.

Good Luck!

Best regards

[Chuck7]

Hi,

I'm also having problems trying to get a natural convection problem to converge. Could you please have a look at my case?

http://www.cfd-online.com/Forums/flu...tml#post539428

Regards

Chuck

[nvarma]

Here are some quick pointers:

1. Run first with Bosinuesq(spell correct :/) and get an initial guess. Then switch to compressible model.

2. If it's a high Rayleigh number flow, change gravity to .098 or something to get an intermediate solution.

did you calculate the timescales of the flow?

also what turbulence model are you using?

[hotboy]

Originally Posted by Bionico

Unfortunately transient simulation would be the best way because the system is intrinsically unsteady. I was able to use those solvers with steady simulation, they took some time to tune up the numerical parameters (CFL, etc...), but they worked well.
If you will have any question about the specific solver I may help you.

Good Luck!

Best regards

Dear Bionico!Thank you very much for your help !I will continue to work hard on it!

[wc34071209]

Hi,

Do you mean coupled solve with pseudo transient enabled can be used to solve steady buoyancy-driven flows?

Does it perform better than SIMPLE?

Originally Posted by Bionico

Hello hotboy,
I worked on natural convection in closed domain: I had the same difficulties as yours and I found useful some solver set up like Density-based solver (Iterations with Solution Steering) or Pressure-based-coupled with pseudo-transient. They are heavier in terms of memory, but they even require less iterations to converge than segregated solvers!

Best Regards