CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Transient Coupled Runner (

pps September 30, 2013 05:26

Transient Coupled Runner
1 Attachment(s)

I have to couple a runner with a stayvanering. The stayvanering is meshed in ICEM, the runner with turbogrid, so I have one mesh as .cfx5 and one as .gtm.

To start a transient simulation I need an initial solution. So i started with an "frozen rotor"-simulation. (stage didn't work)

The first problem with frozen rotor is, that the flow isn't continuous at my interface between runner and stayvanering. (see picture)

The second problem: at the transient simulation my monitorpoint c_p swings too much and is not in the expected range. It also swings with an unexpected frequency.(..swings not with the blade passing frequency)

I think, both problems are result of an bad interface between runner and stayvanering.
Has anybody ideas what exactly could be wrong?


ghorrocks September 30, 2013 18:27

I have no idea what your picture is showing. Can you explain it?

I also cannot see the non-continuous interface. Frozen rotor should be continuous.

pps October 2, 2013 03:35

This is a detail of my runner + stayvane-ring. The only thing I wanted to show is, that at the radius r_m the flow clearly isn't continuous.


ghorrocks October 2, 2013 07:09

A labelled image would help.

Have you tried a finer mesh? Is the GGI connecting over this region? What do the velocity vectors look like in this region?

pps October 2, 2013 07:40

The only important thing in the picture to see is, that in the middle of the colored field, there is a region (red color) that ends abruptly. At the same place is also my GGI Interface.

I've tried a finer(7 million nodes) an a coarser(2 million nodes) mesh, both the same result.

The vectors show a flow, as if the red colored area was a wall. There is also no overlapping at the interface. I checked.


singer1812 October 2, 2013 09:35

Are you using hybird or conservative to display your contours?

I suspect your are displaying conservative. Try display with hybird.

ghorrocks October 2, 2013 17:50

And are you sure this is not real? For instance is it a shock wave or a very steep pressure gradient?

pps October 3, 2013 04:27

I'm using hybrid.

I'm very sure that there is a mistake in the simulation. In physics, there are no jumps, and in this case there also shouldn't be a jump.

ghorrocks October 3, 2013 08:21

I will take you word for that. Have you done the things I suggested in post #4?

pps October 3, 2013 08:40

I also answered in post 5.

..but one question: What exactly do you mean by

Is the GGI connecting over this region?
I configured the Interface at "mesh connection" to connect with GGI. Did you mean that?

ghorrocks October 3, 2013 08:52

Yes you did answer, didn't you :).

A GGI is an interface between two otherwise unconnected meshes. It establishes this interface by initially determining which faces on either side of the interface match up - then the maths starts as it does all the interface stuff across the matching faces. But if it does not match faces up in the first place you can have gaps in the interface which result in weird discontinuities. To check for whether it matched up correctly:
* Look in the output file. It should list the matched faces percentage for each interface. Check this is what you expect.
* Display velocity vectors at each node of a plane through your interface and zoom in to check that the velcoities appear continuous across the interface. If there is a region which has not matched up it should be visible as weird kinks in the velocity vectors.

pps October 3, 2013 09:14

Okay, I'll check that.

pps October 3, 2013 09:18

In the solver-output file i read


Non-overlap area fraction on side 1 = 7,02E-07
Non-overlap area fraction on side 1 = 7,15E-07
Seems to be okay, cause that's ~0.

pps October 3, 2013 10:54

1 Attachment(s)
I looked up the vectors, and there the same problem gets visible:
At many areas the fluid comes through the interface, but at some areas (red colored) there seems to be a wall, where the fluid first has to flow around. :confused:
Of course in real there is no wall..

(maybe better to ignore the added picture. But I think I described the important things.)

I even tried to rebuild the stayvanering-mesh, so that it is nearly similar to the runners-mesh, but I got the same bad results.

pps October 3, 2013 11:35

Could there be any tolerance problems? In ICEM I can set the Triangulation- and Topo-tolerance.

ghorrocks October 3, 2013 18:02

It could be a geometry tolerance. It could also be a fine mesh connecting to a coarse mesh. Can you post an image of the mesh of both sides of the interface?

pps October 4, 2013 04:08

1 Attachment(s)
I also thought about that, but from my view, it seems to be acceptable divergence. ..imagine I would take a tetra-grid.:cool:

I have to ask again: even with frozen rotor there shouldn't be a jump at the interface?

ghorrocks October 4, 2013 05:55

Assuming the interface is the top bit joining the bottom bit and you have separated the two bits for clarity:
Then why have you refined the mesh at the interface? There is no need to do this. Just have your normal volume mesh continue to the interface.

If you have not separated the two bits for clarity:
Then the problem is obviously the gap between the two bits (and the mesh refinement is wrong too).

And to answer your question:
For a frozen rotor interface the flow should not be affected by interface. But note there is a sudden change in frame of reference, so velocities will show a step change. Is this what you are seeing? Try plotting Velocity in stationary frame and this FAQ:

pps October 4, 2013 06:24

oh I'm sorry, i completely forgott to explain my picture.

the picture shows a part of the ring-like interface
the top picture shows the mesh of the runner at the interface, the bottom picture shows the stayvanering-mesh at the interface.

The both Interface-meshes show a refinement (horizontal) at the top and bottom because there is wall. The vertical refinement is because of a blade-wall near the interface.

ghorrocks October 4, 2013 07:12

I see - in that case my previous post #18 is wrong.

Can you post an image of the mesh of both sides of the interface, taken from parallel to the interface. Post #17 shows the view normal to the interface.

All times are GMT -4. The time now is 04:44.