Divergence when restarting simulation

[Anonymous]

Hi,

I'm doing a steady state flow + thermal analysis of flow in a channel. Everything works fine with my simulation until I stop the solver to change turbulence model. In this case I stopped the solver at iteration 400 to switch from standard k-eps to RNG. When I press solve again I get divergence (I also get divergence when I try to resume the solution with standard k-eps). I've tried using both intelligent solution control and large manual under-relaxation.

The flow I'm simulating is pretty extreme and I've had some difficulties with divergence prior to this. I saw in another post that the momentum equations need to be calculated again when resuming a simulation, something which gives a "jump" to the values in the domain. Link:

http://forums.autodesk.com/t5/simulation-cfd/fluctuation-of-residual-plots/m-p/3682238/highlight/tru...

 Could this be the reason it diverges? Is there a way to get around this so that i can resume my simulation?

Any help is appriciated.

With kind regards

Henrik

[Jon.Wilde]

Hi,

This could be down to an issue with the setup.

Also, are you running from a local drive?

Can I ask why the choice of RNG? Are you able to share the model with us at all?

Kind regards,

Jon

[Anonymous]

Ok, I attach the support file.

I believe I'm running it on a local drive. I will look into this; I don't have the computer available presently.

I want to use RNG because its advantage of being more phys

[Jon.Wilde]

Wow, nice - a model where I do not have a single comment to make other than it is very well set up 🙂

Have you considered the SST-kw model at all? It might mean you do not need to make the switch between models.

Testing this here shows that I am able to change from one model to another but only running one iteration of each. 

Thanks,

Jon

[Anonymous]

Hehe 😄 It is however thanks to you and some other support who gave me hints along the way 🙂

Yes, I am doing both k-w and RNG. But I am doing the initial meshstudy with RNG 🙂

Really. With the mesh I had assigned and no divergence? I am running the simulation on C: now(C: is where the program is installed, I did simulations on 😧 before)  and using RNG from start.. just to avoid having to switch models. I guess it will just take a little longer.

I read briefly in another post about running simulations on the cloud for better performance. Is this correct? because right now I need roughly ~48+ for one simulation.

[Jon.Wilde]

OK. I would try SST-kw too.

Sure, the cloud can be faster, it really depends on your local system. At the moment the cloud PC's are 8 core (I think they are around 2.6 GHz but I am not 100 sure).

Thanks,

Jon