[sunset]

I use CFX to simulate centrifugal compressor include one impeller passage and one diffuser passage. There was no problem. But ,when I simulated centrifugal compressor include two impeller passage and three diffuser passage , the CFX solver would encounter error overflow.I don't understand .Can anyone help me?

[ghorrocks]

FAQ: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

[sunset]

Originally Posted by ghorrocks

FAQ: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

I have readed the FAQ. And,I finded the reason about my problem. I modified my mesh many times in these two days. I find some time there was no problem, but I haven't found the law. I talked to many students on the internet these days . I find there are many students have the same problem like me. So, I want to know whether there are some tips or criterion to judge the mesh quality when I use the SST turbulence model. Furthermore, I am not good at english. I am a chinese student. So, please be patient to read my problem. Thank you very much indeed.

[jthiakz]

SST requires Y+ <1 ,so you need better refinement near the wall (prism element). but you cannot just fix it straigt away. bcoz y+ depends on first cell height+shear velocity. So first assume the first cell height of 0.001mm and then review the solution you got. based on that you can improve the mesh further. you can also approximate Y-dist based on flow over a flat plate assumption, check fluent help manual. you can also try to use SST-reattachment modifcation model.

[sunset]

Originally Posted by jthiakz

SST requires Y+ <1 ,so you need better refinement near the wall (prism element). but you cannot just fix it straigt away. bcoz y+ depends on first cell height+shear velocity. So first assume the first cell height of 0.001mm and then review the solution you got. based on that you can improve the mesh further. you can also approximate Y-dist based on flow over a flat plate assumption, check fluent help manual. you can also try to use SST-reattachment modifcation model.

Firstly,thank you for you answer. I have noticed the Y+. But, there is a strange thing that I can simulate my centrifugal compressor with one impeller passage and one diffuser passage without any problem. And I am satisfied with the numericial result when I compare it with the experimental result. However, it will encounter the overflow error when I simulate the centrifugal compressor with tow impeller passages and three diffuser passages under the same setting in CFX. I just duplicate the original mesh along the circumferential direction for the pitch ratio near one. I have no ideal for the reason.

[jthiakz]

# is it still periodic, after duplicating the mesh.
# how about the mass in
# how about the interface between the impeller to diffuser inlet

[sunset]

Originally Posted by jthiakz

# is it still periodic, after duplicating the mesh.
# how about the mass in
# how about the interface between the impeller to diffuser inlet

Yse, it is still periodic.
The interface type is still stage.

[ghorrocks]

SST requires Y+ <1

No, it does not. When used in conjunction with automatic wall conditions it can handle integrating to the wall for y+<1, wall functions for y+>11 and a transition region in between. Do not restrict yourself to y+<1 if you do not need to.

With regard to the dual passage model failing with overflow errors: Have you checked the way the mesh was formed? Is the second passage interfacing correctly with the first?

[sunset]

I opened the CFX-pre and then to build a new turbomachinery case. After the one impeller passage mesh loaded, I edited the Passages and Alignment to choose the Passages to Model with 2. I did the same thing to my diffuser passage mesh. Is there anything wrong with my opperations?

[jthiakz]

"No, it does not. When used in conjunction with automatic wall conditions it can handle integrating to the wall for y+<1, wall functions for y+>11 and a transition region in between. Do not restrict yourself to y+<1 if you do not need to"

I mean to say that,If near wall effects are playing dominant role in your analysis and it is possible for you make such mesh. why to do want to model it with wall function.

[ghorrocks]

Sunset - the thing to do is to look at your mesh and check that the mesh is contiguous over the join. Also check the quality is good.

why to do want to model it with wall function.

If wall functions are adequate then using a finer mesh to achieve integration to the wall is pointless. Only integrate to the wall if you have a clear need to do so.