CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CONVERGE (https://www.cfd-online.com/Forums/converge/)
-   -   How to apply AMR settings (https://www.cfd-online.com/Forums/converge/190845-how-apply-amr-settings.html)

kahjoonyong July 24, 2017 08:51

How to apply AMR settings
 
1 Attachment(s)
Hello,

I am trying to simulate a simple 2D turbulence flow through a channel with hill in order to get my hand on this software. The idea is to be able to hone my skill using this software before proceeding to a simulation for a reactive flow.

The profile is taken directly from this simulation from NASA, https://turbmodels.larc.nasa.gov/Oth..._periodic.html

Here is a link (http://imgur.com/a/hMS7S) to my case setup and geometry.

I have first simulated this case using only fixed embedding at the top and bottom wall boundary. The simulation was done using both k-Epsilon and k-Omega turbulence model. The results could be seen in the link as well. However the base grid I used for this settings is very fine throughout the profile and it took a long time to simulate. Therefore I would like to try out the AMR setting.

However, AMR would not activate throughout the whole simulation. The number of cells do not increase and no mesh refinement was seen in the post processed results. I have tried to coarsen/refine the base grid, manipulate the grid ratio, deactivate Fixed Embedding, manipulate the embedding scale either or both in Fixed Embedding or/and AMR settings, etc. Nothing seems to work. The only time AMR was shown working is when I increase the 'minimum cells' setting to a very high number. Nevertheless, the simulation crashed when the cell count increased.

I have been trying out different combinations of settings over the course of two weeks. I have been looking everywhere in the internet and in manuals, nothing gives useful resources. I hope someone could take a look in it.

I have attached the .cvg file of the case here as well. Attachment 57395
The boundary settings for the wall might vary a little from what I have written in the following lines because I have tested other settings as well. For k-Omega No-Slip boundary conditions are applied.

Short overview of the boundaries:

1. Top and Bottom:
Wall
Velocity: Law of wall
Temperature: Specified value

2. Left Inlet:
Inflow
Velocity in x-Direction 5m/s
Pressure: Zero normal gradient

3. Right outlet
Outflow
Velocity: Zero normal gradient
Pressure: 101325Pa

4. Front and Back
2D

Thank you.

MFGT July 24, 2017 09:45

Hi,

from the CONVERGE Manual:

Definition of "amr_vel_sgs_embed": Sub-grid velocity above which a cell will be embedded. 0.1% to 10% of the characteristic velocity in domain

So i would say if your max. Velocity is somewhat between 5 and 6 m/s, your subgridscale of 1m/s is way to high to trigger AMR.
Try reducing this to maybe 0.25, which would be 5% of 5m/s then.

kahjoonyong July 24, 2017 09:48

Thank you very much to your reply. I will try it out. I have previously tried 0.1 m/s as well as 0.001m/s, both of them failed. I will try again anyway. I will update you as soon as I have a result.

MFGT July 24, 2017 10:12

Which release of CONVERGE do you use?

kahjoonyong July 24, 2017 10:13

I am using Converge 2.3

kahjoonyong July 24, 2017 11:29

Hello MFGT,

I have ran the simulation with a subgrid scale of 0.25m/s. It did not work. Only Fixed embedding is seen in the final results. I have tried to increase the embedding level to 5 in another simulation. It did not trigger.

Blanco July 24, 2017 12:46

Quote:

Originally Posted by kahjoonyong (Post 658227)
Thank you very much to your reply. I will try it out. I have previously tried 0.1 m/s as well as 0.001m/s, both of them failed. I will try again anyway. I will update you as soon as I have a result.

The maximum number of cells set in AMR is 100k, which is too low because your base grid should bring to approx. 142k cells. You should increase the maximum number of cells to allow AMR work properly.

kahjoonyong July 24, 2017 12:47

Hi MFGT,

I have varied the value of the subgrid velocity to an even smaller value and finally AMR triggered! Thank you very much!

kahjoonyong July 24, 2017 12:48

Hi Blanco! That is the reason why it crashed after it triggered! Thank you so much! I just wanted to investigate. :D

ymao July 24, 2017 12:50

Kaihjoonyong,

I have tested this case in Converge.2.3. The AMR actually worked if you changed the sub-grid criterion to very small value, such as 1e-8m/s. At cycle of 501, the cell count increased to 48975 from 14760.

Just for your record,
Regarding on how I choose this sub-velocity criterion of 1e-8m/s, in fact, Converge could output this value through non-transport passive of VEL_SGS.

1) Materials>Species>Non-transport Passive, Add, type VEL_SGS
2) Output/Post Processing > Post variable selection > Cells > Species/Passive, Check box 'Passive', select VEL_SGS through drop-down menu

Given this VEL_SGS, we could test AMR trigger easily.

kahjoonyong July 24, 2017 13:00

Quote:

Originally Posted by ymao (Post 658265)
Kaihjoonyong,

I have tested this case in Converge.2.3. The AMR actually worked if you changed the sub-grid criterion to very small value, such as 1e-8m/s. At cycle of 501, the cell count increased to 48975 from 14760.

Just for your record,
Regarding on how I choose this sub-velocity criterion of 1e-8m/s, in fact, Converge could output this value through non-transport passive of VEL_SGS.

1) Materials>Species>Non-transport Passive, Add, type VEL_SGS
2) Output/Post Processing > Post variable selection > Cells > Species/Passive, Check box 'Passive', select VEL_SGS through drop-down menu

Given this VEL_SGS, we could test AMR trigger easily.

Hello Yijin Mao,

Thank you vary much for your advice! This is going to be very helpful. I will take note of that.

Also shout out to MFGT and Blanco for the quick solution. You basically just saved my day. :D

Kah Joon Yong

MFGT July 28, 2017 05:18

Good to know it finally worked.
Since i use the code for engine simulation only, the amr settings from the examples pretty much fit my simulations.

Also, i cant afford to get more and more cells. The maximum is around 2mio for my cases. So a subgrid scale of 1m/s for velocity and 2.5K for temperature work fine in engine cases.


All times are GMT -4. The time now is 19:07.