I really need some help w/ Memory
Guys, I'm running on fumes. This is for a thesis project. My advisor doesn't even know how to use hypermesh, much less do anything else. I've self-taught myself everything in hypermesh up to this point. And this is not even the smallest mesh I'm going to have to produce and my computer is crapping out. Can someone please help me get this resolved? I do have the ability to run the simulation on a supercomputing center, but it is linux based and I have to somehow convert this to Openfoam. I have no clue where to begin. This forum is really the only reason I've gotten this far and seems to be the only reliable medium of assistance that I can get. I've tried core out, the param route with checkel=no and hashassm=yes, the tmpdir fix, and I'm still not getting anything. And I know that have more than the available memory to run this, because I'm watching my PC performance as the sim runs. Please help.
Answers
-
First things first:
you have a huge model. with 24M elements. This is A LOT!
Are you sure you need a so detailed model??
A full car model would have 1~2M elements.
FINITE ELEMENT MODEL DATA INFORMATION :
---------------------------------------Total # of Grids (Structural) : 4666863
Total # of Elements Excluding Contact: 24802640
Total # of N2S Contact Elements (3D) : 574948 (internally created CGAPG)
Total # of Degrees of Freedom : 15725433
(Structural)
Total # of Non-zero Stiffness Terms : 406563360Your model requires a MINIMUM of 18860 MB RAM for running. If you could at least free around ~20GB it might start running out-of-core.
But again, I would make sure you need this big model.
Recommended Memory (RAM) for Out of Core Solution : 18860 MB
Recommended Memory (RAM) for In-Core Solution : 45321 MB0 -
/profile/100300-awe901/?do=hovercard' data-mentionid='100300' href='<___base_url___>/profile/100300-awe901/' rel=''>@awe901: show us some details of your hardware?
24M elements model is big. A big model needs a 'big' hardware to solve.
OpenFOAM is Open Source software, so you have everything on Internet free to learn
/emoticons/default_biggrin.png' srcset='/emoticons/biggrin@2x.png 2x' title=':D' width='20' />
Linux OS is also Open Source software, you can learn it no issue. It's my favorite OS
0 -
The golden rule of finite element analysis: An FE model should be as simple as possible, but as complex as necessary.
Start with a coarse mesh and decide on the quantity of interest (usually displacements or stress, with the former converging faster than the latter), then solve it using progressively smaller mesh size (mesh convergence study), reducing mesh discretization error.
https://enterfea.com/correct-mesh-size-quick-guide/
If you plan to do a convergence check, consider performing at least one refinement of the model after the first run. If neighboring elements display large differences in quantity of interest, the gradient was probably not captured in these areas, therefore some mesh refinement is recommended. In general, increasing the number of nodes improves the accuracy of the results. But at the same time, it increases the solution time and cost. Usual practice is to increase the number of elements and nodes in the areas of high stress (rather than reducing the global element size and remeshing the entire model) and continue until the difference between the two consecutive results is less than 5 to 10%.
I recommend you to go through free ebook Practical Aspects of Finite Element Simulation (A Study Guide) which covers 9.12 Mesh Density And Solution Convergence.
0 -
Altair Forum User said:
Guys, I'm running on fumes. This is for a thesis project. My advisor doesn't even know how to use hypermesh, much less do anything else. I've self-taught myself everything in hypermesh up to this point. And this is not even the smallest mesh I'm going to have to produce and my computer is crapping out. Can someone please help me get this resolved? I do have the ability to run the simulation on a supercomputing center, but it is linux based and I have to somehow convert this to Openfoam. I have no clue where to begin. This forum is really the only reason I've gotten this far and seems to be the only reliable medium of assistance that I can get. I've tried core out, the param route with checkel=no and hashassm=yes, the tmpdir fix, and I'm still not getting anything. And I know that have more than the available memory to run this, because I'm watching my PC performance as the sim runs. Please help.
Hey Awe901,
What do you want to do whit 15 M DOFs ? I'm just curious ... Are you trying do perform Rough contact simulation ? Damage simulation ?
0
Altair Employee