CFX Diffuser Simulation
 

Hi,

I have some further questions relating to a previous post with the same title (see a few threads back).

I'm using inlet BC - "Opening" with "opening pressure and direction" set to zero and "perpendicular to boundary" respectively as this gives the best convergences (studying grid dependance currently). Also for the same reason outlet set to "outlet" and "mass flow". I've tried Mass flow at inlet but this results in oscillating residuals above 1e-04, the current setup allows them to drop to under 1e-05 where I stopped the run. They were still descenfing at this point.

My questions;

1) Would I be correct in thinking that as the flow is almost certainly unsteady, the "better" convergence I get with this particular set-up is just due to slightly more robust BCs, rather than my simulation being any more accurate or "better"?

2) Can anyone explain why despite having Y+ 80 (using KE), zero slip wall condition etc, at the inlet I have highest velocities adjacent to the wall? The boundary layer develops further in, but there is a distinct low velocity core on the inlet boundary (approx 26 m/s as opposed to 34m/s adjacent to wall? This looks questionable to me but I have no explanation as yet.

Thanks!

Ianto

Re: CFX Diffuser Simulation
 

Hi,

Q1 - To check whether the result is transient you need to do a transient simulation. The issues regarding mass flow inlets and outlets is discussed in the CFX documentation, in obtaining convergence.

Q2 - Could it be related to temperature near the wall somehow? Do you have something close to the inlet which could influence the flow at your inlet?

Glenn Horrocks

Re: CFX Diffuser Simulation
 

Thanks Glenn,

I need to re-read the documentation. I usually need to reread everything several times!

I'm not using any thermal simulation at all as I thought this may help convergence initially, I haven't got as far as turning on isothermal - this is the furthest I intended to go as I don't think thermal effects were particularly significant in water at 28 m/s....... Would you agree? THere is nothing in the inlet vicinity to cause any problem I don't think. I'm modelling it as the flow passage only, with one inlet, one outlet - no other items in or adjacent to the domain at all.

I'm currently rebuilding meshes as the mesh dependance study showed some cases of increasing difference in calculated pressure drop with refined mesh which seems odd, I did have some warnings when meshing which I hoped I could ignore, but now I guess this should be the first area to direct my attention since the solution does not converge well and the differenc in pressure drop gets bigger with refinement. ANy comments gladly received!

One problem which I'm having with the CAD model/mesh is shown below (the warning message) below.. I can't find any visible anomaly in the models, and have tried various virtual face/edge combinations as well as saving the Solidworks model as aboth IGES and parasolid and reimporting. The edge which this error refers to changes also which is puzzling me. Again, any suggestions would be very welcome!

Many thanks,

Ianto.

Warning Problem The distance grid nodes move from a CAD edge to a CAD face exceeds 1.000E+01% of the CAD model accuracy. (This occurs 23 times.)

ds = is the distance moved from an edge to face.

Max dist = 1.812E+01% [100*ds / e_len] Mesh spacing (e_len) 1.507E-03 Distance moved (ds) 2.731E-04 CAD tolerance (tol) 1.235E+00

Result This may result in a poor quality volume mesh or failure of the surface mesher.

Cause If the collapse face/edge option is active then an attempt has been made to collapse a CAD face or edge which is of a similar size to the local mesh spacing.

Action Increase mesh spacing in region of face.

Action Do not remove sliver faces or short edges.

Cause CAD model tolerance is of a similar size to local mesh spacing.

Action Fix CAD model or increase local mesh spacing.

Re: CFX Diffuser Simulation
 

Hi,

That warning message suggests your geometry is complicated and that means generating a good quality mesh could be hard. Poor quality mesh can cause strange issues, is your mesh good quality?

Glenn Horrocks

Re: CFX Diffuser Simulation
 

Hi Glen,

I interpreted the message as an indication the CAD model was poor, resulting in problems for the mesher trying to tie the part together. However, I couldn't find any obvious defects in the CAD model (to be fair this is the first project so not entirely sure what I was looking for - assumed slivers, non-coincident end points, gaps in geometry and so on). It's a quite complex shape - a diffuser opening into a strongly 3D 180 degree curved passage, which I have mixed success in meshing - some models I have no errors or warnings from the mesher, others I manage to get 30 < cell aspect ratio < 50 at best.

It was suggested in a previous post that the limits advised by the software are quite conservative and can be exceeded some way without adverse effect, though obviously I'd prefer to stay within them.

I have Y+ of 80 for KE and 130K, 240K and 340K cells to determine grid independence.

Thanks & regards,

Ianto

Re: CFX Diffuser Simulation
 

Hi Ianto,

I have heard from an experienced friend of mine of relatively similar problems. What I suggest to do (if you have not done it already) is to create (grow) several layers of prisms starting from the walls into the fluid that you are using. I have had success using this approach.

Best regards, Alin

Re: CFX Diffuser Simulation
 

Hi Alin,

Thanks for your help, I did use this approach already. In CFX it's called an inflated layer - you increase the mesh resolution adjacent to the wall to better resolve the viscosity affected boundary layer. Velocity gradients are steep here (as there is zero slip at the wall) hence the need for smaller mesh spacing to capture the flow physics. Spacing is smallest immediately next to the wall, and increases with each layer away from the wall, hopefully smoothly blending with the bulk mesh spacing (I have probs with this just now!). You should use a Y+ value (non-dimensional parameter corresponding to distance from wall to first cell centre) which is appropriate for the turbulence model employed (see software user manual for guidance). This varies depending on turbulence model, whether you are using wall functions, low Reynolds version of the turbulence model (or one of the automatic options available in CFX). Best advice is refer to the user manual for guidance on Y+ values , and to Schlichting "Boundary Layer Theory" for details on the physics and why this is necessary!

Best regards,

Ianto.