CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   STAR-CCM+ (https://www.cfd-online.com/Forums/star-ccm/)
-   -   The best way to improve mesh (https://www.cfd-online.com/Forums/star-ccm/139132-best-way-improve-mesh.html)

djordje8 July 17, 2014 13:59

The best way to improve mesh
 
Hi,

I created a mesh with 3.5 million cells for my simulation. I ran the job, but now, I would like to run the same job, but to increase the number of cells to around 5 millions. So, I did the following:
1. I opened the project (that has 3.5 million cells),
2. Increased the base size of cells,
3. Initiated meshing, and,
4. Started generate volume mesh

So, I didn't clear previously created mesh. The problem is that it takes forever to create this mesh. It took maybe 30 min to create 3.5 million cells mesh and after almost a day it didn't finish this new mesh.

What would be the best (most efficient) way to increase number of cells and then to re-run the job?

Thanks,
Djordje

me3840 July 17, 2014 23:40

How much did you change the base size by? If you change it dramatically you can easily blow your cell count up. Was your machine swapping during this process? If you're running out of memory I can see where it would take a day to mesh something small.

djordje8 July 19, 2014 08:24

The problem is that whether I increase or decrease the mesh size it takes forever to create mesh (and it doesn't finish). For example, my domain is about 1.5 m tall and several meters long. So, just to make a test, I set the base size to 1 m and costume size in the my area of interest to 20 cm (horrible mesh, but it should take a few second to generate, I guess). I worked almost 1 day and it didn't finish (not giving me an error message either).

Also, if I increase/decrease base size for a reasonable amount (a few percents) it again takes forever to finish. It took 30 min or so to create the original 3.5 million cells mesh.

Do I have to clear the solution created with the previous (3.5 million cells) mesh in order to create a new mesh? What could be a problem?

me3840 July 20, 2014 22:22

Is this a trim mesh?

djordje8 July 21, 2014 09:26

It's Polyhedral Mesher with Prism Layer Mesher and Surface Remesher.

OK. So I am running simulation for an urban block. I have scaled the geometry to 1:400 to be able to run it in starccm+. It didn't allow me to have computational domain larger than a 1 km or so. I have created the volume around the urban block where the mesh (absolute, costume) size is 0.005m. The base size in the rest of the computational domain is 0.1m. I have 4 prism layers outside of the block and 5 in the urban block. This configuration gave me around 3.5 million cells. I ran the simulation and now I just want to re-run it with 5 million cells mesh.

When I change the base size, custom size, number of prism layers, or anything it takes forever to create the mesh.

me3840 July 21, 2014 19:28

Something is missing. Is your machine running out of RAM during the meshing process? Where does it get stuck at?

ggulgulia July 22, 2014 06:04

Hello

I hope that you have sufficient virtual memory in your system. STAR CCM+ needs around 1.3-1.5 gb per million polyhedral cells and around 1 gb per million of trimmer cells. More of memory is required for solving the problem. If you are getting stuck somewhere, it might be due to problems in surface mesh.

I suggest you first load surface remesher model only without any volume meshers. Under the surface remesher model option, disable automatic surface repair option and change the 'Face Quality' value to 0.2

Once done, I suggest check the remeshed surface in surface repair option and check for 'Pierced Faces' and 'Face Proximity'. You MUST make sure that there are NO pierced faces and try to minimize the number of 'Face Proximity'

After this again go to mesh models and select your volume meshing models and remove the surface mesh model. (On hitting the volume mesh, CCM+ runs the surface remesher model too it it's present)

djordje8 July 27, 2014 12:17

I found out what the "problem" was... It's actually embarrassing to say. When I went to vacation, the colleague scaled the model to much bigger size, but the mesh parameters were the same as for my small domain.

Thank you so much.


All times are GMT -4. The time now is 21:06.