Learning Forum

[enie.dreyling]

Hello,

iam trying to simulate particle gas streams with interaction of the particles with a vol fraction<10%.
So far ive used the DEM model but somehow various errors occur when running my simulation. 
Usually they are AMG solver errors for epsilon, k, pressure and i dont know if i should use another solver. 
Iam currently using the pressure based solver, since im using argon and ive tried with the SIMPLE and Couped solver.

Id be very grateful if anyone can tell me more abou the DEM model or the AMG Errors.

Thank you!

[Rob]

DEM tends to require a very small time step due to the way the particle collisions are modelled (read the Theory Guide). With a low volume fraction look at DPM. Stability issues can also be caused by poor mesh and/or models, boundary conditions etc. Have you sanity checked the settings and reviewed the cell quality? 

[enie.dreyling]

Thank You for your reply.

I will try an even smaller time step (1e-04?)

As i need the interaction in my case i cannot use just the dpm model.

I tried everything i could with new boundary conditions and they should be fine, the mesh quality is good aswell.

My simulation ran once for 50iterations with first oder discretization, but when i tried second order momentum etc. it diverged and the AMG Error occured again.

 

[enie.dreyling]

 

Thank You for your reply.

I will try an even smaller time step (1e-04?)

As i need the interaction in my case i cannot use just the dpm model.

I tried everything i could with new boundary conditions and they should be fine, the mesh quality is good aswell.

My simulation ran once for 50iterations with first oder discretization, but when i tried second order momentum etc. it diverged and the AMG Error occured again.

 

[Rob]

Yes, read up on the particle collision model, you need to resolve the collision time. 

[enie.dreyling]

Okay, i resolved the time stepping but one error still occurs which is: turbulent viscosity ratio limited to ... in cells.

This once came up before even without particles and just argon. Is this also a solver problem because for my particles i thought i could use the default SIMPLE and not Coupled.

Thank You for your help!

[Rob]

Check the turbulent values on the boundaries and how well velocity gradients are resolved. Chances are you've missed something in the set up unless argon has some weird density & viscosity values. 

[enie.dreyling]

I got the model to work, now i am trying to alter the number of parcels which are being injected every time step, because they are quite a lot and require too much memory.

I have read the users and theory guide but whenever i change the parcels constant mass no parcels are injected at all anymore.

Does this have to do with adaptive time stepping or am i missing something else?

Kind regards

[Rob]

Messing with the parcel mass options can be a little complicated, not least as if you get it wrong you'll have no injections (typically because the injected mass per parcel is too small). Have a look at injection files and/or other options for a surface injection: group might do what you want. 

[enie.dreyling]

Thank you for your help!

I got the model working to some point with a time step of 1e-06 but some problems still occur.

I was thinking that first order discretization for turbulence and momentum could be better than second order? Because at the moment iam using second order and epsilon is not converging. 

Thank you again 

[Rob]

Given turbulence is an empirical model with some level of approximation/assumption I do tend to run first order for turbulence on many models. Have a look at Higher Order Term Relaxation (HOTR) too.