[surajkashyap]

Hi,

I am modelling the volute of a centrifugal pump in ICEM-CFD.
Upon running a quality check (determinant 3x3x3), I get some elements which have a quality of less than 0. I have tried a lot, but I cannot fix these elements without messing up others elsewhere.
However, these poor quality elements are just 29 in number out of a total of ~459000 in the entire model (~0.02%). When I run the model in CFX solver, I do not get any error messages, and the output also matches experimental values quite fine.
Is it that I am missing something here? Is it possible to have a few elements of poor quality and still have a good solution?
I would have gone forward satisfied with the results but this is for my Undergrad thesis, so I do not wish to take the risk.
If anybody can take a look at the model and help me fix these last few elements, it would be great! Thanks

* I am unable to attach the .zip file here as it exceeds the size limit, but if anybody is willing to take a look, I can mail it to them.


Update : I do get error for negative volumes in CFX Solver when I push up the refinement and the number of elements goes up to ~3000000.

[diamondx]

Refinement make cells smaller. And from my experience this always gives negative volumes. Even in icem CFD. What is the problem ?? Having bad quality does not always implies bad result. Those elements are not of an interests for the flow sometimes. Scalars may be of a bad value bad overall flow can be good.

[surajkashyap]

Originally Posted by diamondx

Refinement make cells smaller. And from my experience this always gives negative volumes. Even in icem CFD. What is the problem ?? Having bad quality does not always implies bad result. Those elements are not of an interests for the flow sometimes. Scalars may be of a bad value bad overall flow can be good.

@Mr. Ghazlani Thank you for clearing that up. I am new to CFD and was under the impression that one needs to have a perfect histogram, with determinant > 0.1 or 0.2, as supported by the solver.
As a follow-up question, would having a bad mesh compromise the credibility of my work for purposes such as publishing papers or participation in conferences, even if I validate my work with experimental values? Apologies if my questions sound very inexperienced.

[Far]

You should go for the quality > 0.3 and angle > 18. Yes poor quality means poor results and means no publication. Improve your mesh. This is the most difficult case for ICEM CFD Hexa.

See Simon's presentation for handling same geometry in 2d.