2012 CFD Simulation Wont Finish

[timsefton]

Working on getting a CFD model of a simplified air amplifier complted - got the model in and meshed and analysis running but it wont complete - gives me a message tha anslysis didn't finish but there are no errors - any thoughts.

Here is the mesh - with inlet/outlets at nozzel end and back opening - and an external intake fan at inlet nipple (orange).

Here is a section view:

I'm using a load curve as follows:

Any ideas greatly appreciated !!

Tim

[Joey.X]

There are lots of factors affecting simulation convergence/result. According your pictures, first check following

(1) If your fluid simulation domain is on fluid domain only; if not, you need create fluid domain in CAD, and disable solid parts in simulation.

(2) Check internal fluid flow Reynolds number(Hand calculation). For internal flow, if Re>2300, the turbulence option has to be turned on in load curve table.

(3) Check mesh density, The too coarse mesh density in some thin part may block flow numrcially and introduce simulation divergence.

[timsefton]

Hi Joey,

Thanks for the ideas,  one quesiton, not sure what you mean when you say:

"If your fluid simulation domain is on fluid domain only; if not, you need create fluid domain in CAD, and disable solid parts in simulation."

I'm creating an internal fluid in Autodesk Simulation from defining the interior surface and boundary surfaces - then meshing all three parts (insert, casing and internal fluid) excluding the boundry surfaces from boundry layer and then analysising.

Not sure what you mean by creating fluid domain in CAD - I do see some velocity vectors on the exterior of the parts when it stops in the analysis mode which indicate to me I have the exterior surfaces involved with the interior fluid flow analysis, which is obviously incorrect.

Still Learning,

Tim

[Joey.X]

Hi, TIm,

To clarify, the fluid simulation works on fluid parts only. In other words, if solid parts involve in fluid simulation for some reason( or modeling mistake), the unwanted wrong result or solution divergence may introduce.

It appears that the parts are void in your 2nd picture while the inlet and outlet BCs are applied in the void part. And the mesh in 1st picture is on the exterior surfaces of solid part.

For the second question, if you create fluid part in simulation, make sure the wanted part has been really created, and disable other solid parts before doing fluid simulation.  If creating fluid part is not successful, I suggest creating fluid part in CAD using boolean operation which sounds a more robust way to me. 

[timsefton]

Thanks Joey,

Yes that makes sense - I did go ahead an disable the other parts but something is still out of wack - in one instance I did get some partial results but no completion - what is frustating is the almost complete lack of feedback from the application when it does fail -  makes debuging very diffcult.

One more question, if I define a external intake fan on a surface, does that surface need to be defined as an inlet / outlet also or is that taken for granted ?

Tim

[Joey.X]

(I did go ahead an disable the other parts but something is still out of wack - in one instance I did get some partial results but no

completion)

- Check mesh setting (mesh density in critical location as mention above,  and boundary layer mesh)

- Try unsteady analysis with proper time steps

(One more question, if I define a external intake fan on a surface, does that surface need to be defined as an inlet / outlet also or is that taken for granted ?)

- No, fan BC can't be overlapped with inlet/outlet on same surface, it's over-constrainted. 

[timsefton]

Thanks,  I'll give it a shot.

Any suggesitons on time steps for load curve ?

Tim

[Joey.X]

For time step

(a) In steady fluid flow analysis, the time step is pseudo and it has no effect to result.

(b) In unsteady analysis, it suggested starting from relative small time steps (the value depends your model's nature) to get satisfied result, and then adjust time step to larger one. 

For load curve, just set proper load curve magnitude for the loading you want, it's usually not necessary to set magnitude progressively for convergence consideration, solver already takes care of this by using "inertial term" in governing equation.

[timsefton]

The simulation is still not completing - tried both steady state and tranisient simulations -

For steady state, I get the will not finish along with a note in the analysis log -

Non-positive diagonal entry detected in ILU method swith to SSOR preconditioner and then it goes a bit further and stops - does give me some type of results:

For the unsteady fluid flow  with the following load 

I get the following:

I thought it would be a simple anaysis - one fan input and two outlets - 

Any thoughts ?

Tim

[timsefton]

Here is what the mesh looks like if thats any clue -

[Joey.X]

- Check inlet fan BC, prefer flow rate or velocity based fan BC which are more robust to solution convergence.

- Check mesh density in narrow channel at right side(Message 9 of this thread), refine mesh or use boundary layer mesh if too less interior nodes in the narrow channel.

[hupn]

Hi,

For transient fluid flow setting, if your solution is already stuck at the beginning, you may try to set small multiplier initially and increase gradually (you have all 1 currently and seems to have convergence issue at the beginning already). Try use BCSLIB_EXT solver to see if you get a better convergence, thought it will take longer to solve. Generally, the convergence issues show that the system is not in good condition, either the problem settings, or the mesh, or some other issues.

hupn   .

[Anonymous]

I am having problems getting the problem to finish. It does not give a reason for stopping. It goes for 1000 iterations.

What can I try?