Laminar_turbulent_transition. Tetra or hexa?

Roland R · November 21, 2012, 10:09

Hello,

I complated two 3D (steady state) simulations to compare the influence of tetra and hexa mesh in case of calculation of laminar-turbulence transition. The SST turb.model and the gamma-theta transitional turbulence have been used.

Based on results, the y plus is same but the location of transition is different. The turb.kin.energy is higher in case of hexa mesh.

The boundary layer contains at least 12 elements and I tried to keep all of requirements in terms of meshing.

This simulation was just a test with a simple 3D geometry. I have to investigate the transition in case of a quite complicate geometry where it's likely that I can not apply structured hexa mesh. Therefore I would like to know whether it is possible to calculate the transition exactly with tetra mesh?

Regards
Roland

cdegroot · November 21, 2012, 12:25

A hexa mesh is in general better than a tetra mesh. However, I believe that with sufficient refinement (could be a LOT of refinement) it should be possible to get the same results out of a tetra mesh. In the limit of infinite mesh refinement the disretization error goes to zero, so it shouldn't matter about mesh type. That being said, you can't actually refine infinitely

Keep refining your tetra mesh and I would expect it would converge upon the hexa result (assuming those results are sufficiently grid independent).

ghorrocks · November 21, 2012, 19:06

First of all, good work on checking this first. It is important to know what the models limitations are. To many people on this forum seem to jump in and wonder why their model is inaccurate..... <END RANT

>

It is difficult to achieve full mesh refinement with the more complex models, such as transition, multiphase, chemistry. Chris's comment is correct, you may just need to go finer again, but often you hit limitations of available resources and you just have to accept that your model has an error in it. So try to quantify the error if possible so you can put a reasonable estimate of error on the results.

Roland R · November 23, 2012, 06:34

Hello,

thanks for your answers. I generated a very fine tetra mesh for my simple model yesterday. The result is better than the previous one. Of course there are problems and questions in the future too

Please help me to understand the maining of the "turbulence intermittency" and the "transition onset reynold number" variables. I read the cfx help, but I dont understand these wholly.

By the way, my main problem is that it is very difficult to generate the proper transition between the prism and the tetra elements. I try to attend the mesh expansion factor in spite of this the mesh transition and the mesh quality will not be the best...

Regards
Roland

ghorrocks · November 23, 2012, 06:47

There is some very good papers describing the turbulecne transition model - if you cannot find them in the literature then the CFX community site (on www.ansys.com) might have it, otherwise contact CFX support.

But in short transition onset Re number is the variable which keeps track of the Re number the flow will transition to turbulence given its history - smoother flows have higher Re transition numbers. Unfortunately it has been too long since I used that model and I cannot remember what the turbulence intermittency variable is -hopefully somebody can fill in for my memory.

Consider doing Richardson extrapolation to assist in the mesh refinement. It can be very useful in assessing whether you are in the convergent region of mesh refinement, and if you are it can extrapolate to the idea mesh. A very useful tool.

There are lots of meshing tutorials about and hopefully some of them cover the prism to tet transition. Again look on the web, the ANSYS community site and CFX support.

aledexe · November 23, 2012, 10:56

If I'm not wrong, intermittency (gamma) is the probability that a "point" is within a turbulent region..

ghorrocks · November 25, 2012, 05:06

Excellent, thanks for the clarification!

brunoc · November 26, 2012, 07:48

In addition to the y+, you should also check that mesh sizes over the wall in the flow direction are similar for both meshes. That's because mesh refinement in the flow direction is very important for the transition model since it affects how well it can calculate the transition position. The documentation has some guidelines on that.

I'm just saying this because when people create hexa meshes they usually take advantage of the higher aspect ratios allowed by hexahedrons, which could affect the results if you have sizes that differ from the ones in your tetra mesh.

November 21, 2012, 10:09	Laminar_turbulent_transition. Tetra or hexa?	#1
Roland R Senior Member Roland Rakos Join Date: Mar 2009 Posts: 131 Rep Power: 17	Hello, I complated two 3D (steady state) simulations to compare the influence of tetra and hexa mesh in case of calculation of laminar-turbulence transition. The SST turb.model and the gamma-theta transitional turbulence have been used. Based on results, the y plus is same but the location of transition is different. The turb.kin.energy is higher in case of hexa mesh. The boundary layer contains at least 12 elements and I tried to keep all of requirements in terms of meshing. This simulation was just a test with a simple 3D geometry. I have to investigate the transition in case of a quite complicate geometry where it's likely that I can not apply structured hexa mesh. Therefore I would like to know whether it is possible to calculate the transition exactly with tetra mesh? Regards Roland

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Solution on hexa mesh vs solution on tetra mesh	Chander	CFX	2	December 10, 2011 10:35
unstructured or Structure Cartesian grid vs tetra and hexa	eleazar	Main CFD Forum	1	March 21, 2011 18:02
Both Hexa & Tetra to the same Geometry?	Jaloha	CFX	3	April 28, 2008 06:46
Tetra or Hexa	eric	CFX	3	January 31, 2007 23:23
HELP! TETRA and HEXA boundary	sunlight007	Siemens	6	August 6, 2004 07:15

November 21, 2012, 12:25		#2
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	A hexa mesh is in general better than a tetra mesh. However, I believe that with sufficient refinement (could be a LOT of refinement) it should be possible to get the same results out of a tetra mesh. In the limit of infinite mesh refinement the disretization error goes to zero, so it shouldn't matter about mesh type. That being said, you can't actually refine infinitely Keep refining your tetra mesh and I would expect it would converge upon the hexa result (assuming those results are sufficiently grid independent).

November 21, 2012, 19:06		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	First of all, good work on checking this first. It is important to know what the models limitations are. To many people on this forum seem to jump in and wonder why their model is inaccurate..... <END RANT > It is difficult to achieve full mesh refinement with the more complex models, such as transition, multiphase, chemistry. Chris's comment is correct, you may just need to go finer again, but often you hit limitations of available resources and you just have to accept that your model has an error in it. So try to quantify the error if possible so you can put a reasonable estimate of error on the results.

November 23, 2012, 06:34		#4
Roland R Senior Member Roland Rakos Join Date: Mar 2009 Posts: 131 Rep Power: 17	Hello, thanks for your answers. I generated a very fine tetra mesh for my simple model yesterday. The result is better than the previous one. Of course there are problems and questions in the future too Please help me to understand the maining of the "turbulence intermittency" and the "transition onset reynold number" variables. I read the cfx help, but I dont understand these wholly. By the way, my main problem is that it is very difficult to generate the proper transition between the prism and the tetra elements. I try to attend the mesh expansion factor in spite of this the mesh transition and the mesh quality will not be the best... Regards Roland

November 23, 2012, 06:47		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	There is some very good papers describing the turbulecne transition model - if you cannot find them in the literature then the CFX community site (on www.ansys.com) might have it, otherwise contact CFX support. But in short transition onset Re number is the variable which keeps track of the Re number the flow will transition to turbulence given its history - smoother flows have higher Re transition numbers. Unfortunately it has been too long since I used that model and I cannot remember what the turbulence intermittency variable is -hopefully somebody can fill in for my memory. Consider doing Richardson extrapolation to assist in the mesh refinement. It can be very useful in assessing whether you are in the convergent region of mesh refinement, and if you are it can extrapolate to the idea mesh. A very useful tool. There are lots of meshing tutorials about and hopefully some of them cover the prism to tet transition. Again look on the web, the ANSYS community site and CFX support.

November 23, 2012, 10:56		#6
aledexe New Member Alessio Join Date: May 2012 Posts: 9 Rep Power: 14	If I'm not wrong, intermittency (gamma) is the probability that a "point" is within a turbulent region..

November 25, 2012, 05:06		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	Excellent, thanks for the clarification!

November 26, 2012, 07:48		#8
brunoc Senior Member Bruno Join Date: Mar 2009 Location: Brazil Posts: 277 Rep Power: 21	In addition to the y+, you should also check that mesh sizes over the wall in the flow direction are similar for both meshes. That's because mesh refinement in the flow direction is very important for the transition model since it affects how well it can calculate the transition position. The documentation has some guidelines on that. I'm just saying this because when people create hexa meshes they usually take advantage of the higher aspect ratios allowed by hexahedrons, which could affect the results if you have sizes that differ from the ones in your tetra mesh.