Learning Forum

[m.hamza]

Hi,

 

I am trying to mesh in AIM for a fluid-solid heat transfer simulation. I have 64GB RAM on my test machine. When I start the meshing process almost 63.5 GB of memory is used up for the first hour or so (which is good). But then the memory usage drops down to about 14-16 GB only. The meshing task been running with this reduced memory consumption for about 3-4 hours now. Is there any setting that I can tweak to get it to continue using all of the available memory, and hopefully finish the meshing task faster?

 

Thanks!

[Gaurav Sharma]

Hi mhamza

I looked into the database to see if there is some setting which may help you but unfortunately, I could not locate anything that can optimize the CPU usage to that level. The best you can do is selecting Use all Processors = Yes under mesh settings. This is also the default option so I think you are already getting the best response, unless this option is set to NO at your end.

Possibly, most of the resources are utilized initially in meshing different parts of your model, and as the meshing is completed on some of the parts, you can see the memory is getting relieved.

[m.hamza]

Gaurav_ANSYS Thanks for the reply Gaurav. Yes, I do have the same settings to use all the processors.

I was also looking around and I found two settings about which I am not sure what they exactly do.

 

1. Tools > Options > Mechanical APDL > Database Memory (Default value = 1024 MB)

2. Tools > Options > Mechanical APDL > Workspace Memory (Default value = 2048 MB)

 

Since these values are so low, I assume they are not the cause of my problem anyways, however I am still curious as to what it does.

[Gaurav Sharma]

mhamza 

Mechanical APDL memory is divided into two blocks: the database space that holds the current model data and the scratch space that is used for temporary calculation space (used, for example, for forming graphics images and by the solvers). The scratch space is the total workspace minus the database space. 

These display values (Database and Workspace memory) are low but they are just initial allocations, program dynamically allocates the right memory, trying to make maximum use of the available hardware. But, if you will change these values manually, it may end up being the upper limits for program. Hence, it is generally not recommended to tweak these values.

I believe the problem you are dealing with has a bulky/complex geometry and it might not be possible to bring down the meshing time.

[m.hamza]

Gaurav_ANSYS Thanks for the clarification!

[m.hamza]

Dear  Gaurav_ANSYS , I am indeed struggling with getting the mesh settings right for my model. After multiple days of running, something or the other always seems to go wrong. Would you be so kind as to please take a look at my model and suggest the right / appropriate settings for meshing? I am trying to solve for fluid-solid heat transfer in a shell and tube heat exchanger. I am using hex elements for 2 out of the 3 sim domains. Do you think this is right? Please advise on proper mesh settings. 

I would really appreciate your help in this regard. I am trying to convince my team to switch to AIM from our current CFD pacakge and efficiently solving this task would go a long way in making a strong case for the switch...

 

Thanks and Regards

[Gaurav Sharma]

mhamza 

I had to remove the file location form forum as its a public forum. You can send the files to ansys-discovery-support@ansys.com and I will pick it up from there.

Also, please elaborate a bit more on the problems you are facing with meshing, is the meshing failing when you say something goes wrong? What do you imply by "Struggling with getting the mesh settings right".

[m.hamza]

Gaurav_ANSYS I understand. I have sent you another mail to the address you provided. 

Thanks for your help.

[Gaurav Sharma]

mhamza 

I have got the files. Will into them and get back to you.

[m.hamza]

Gaurav_ANSYS Fantastic!

[Gaurav Sharma]

mhamza 

I looked into the model and can see that the model is quite complex to be hex meshed and thats the reason you are unable to proceed. My suggestion is what you have already mentioned, as a next step please try with tet mesh. I am sure that is achievable as I tried with default tet mesh and could get the meshing done with an expectedly higher mesh count. But I guess, considering the complexity of model, you will have to live with it. 

Regards,

Gaurav

[m.hamza]

Gaurav_ANSYS Thanks for looking into this. You were a great help. Very good service! Wish you a great weekend ahead.