Cancel
Showing results for 
Search instead for 
Did you mean: 

No sufficient memory

Builder
Builder

Well, 

for the huge model I have, the nonlinear analysis did not succeed to run due to Ram memory problem.

Now, I switched to Linear analysis but still have the same problem. it seems the glueing is the source of the memory problem.

 *** USER FATAL MESSAGE 3000 (SITDELC)
INSUFFICIENT MEMORY SPECIFIED FOR THE ITERATIVE SOLVER. INCREASE
THE MEMORY LIMIT BY 13343 MB
Noted that the allocated memory was 16GB.

Does that mean model requires 13.4+16GB memory space?!

in running the model, the analysis.exe used no more than 12GB. I assume that this what I need, correct?

I noticed that I was using the default Nastran solver in the environmental variable which has reported here is not suitable for higher memory. I am running now the Nastran L64 with the memory of 15.9gb. Did not worked as well

 

is there a better than trial and error method to solve these problems, please?

7 REPLIES

Re: No sufficient memory

Legend
Legend

Is there a way you can share you dat file? I don't mind running the SOL106 & 101 for you to see what's up...

 

Is it NXN11 that you're running?  I'm asking because ILP64 is the default with NXN11, so you wouldn't need to change anything...  The iterative solver runs in RAM, and it looks like it failed as it was filling it up, and wanted more but couldn't get it...  How big is this model???

Re: No sufficient memory

Builder
Builder

Hi 

 

The dat file is 2.5GB, The model part files compressed about 75MB. They cannot be uploaded here .

If you send me your email will be glad to send the files to you.

 

BTW, I am using NXN10.2 

Thanks

 

Re: No sufficient memory

Legend
Legend

Wazy wrote:

Hi 

 

The dat file is 2.5GB, The model part files compressed about 75MB. They cannot be uploaded here .

If you send me your email will be glad to send the files to you.

 

BTW, I am using NXN10.2 

Thanks

 


Can you post that to a dropbox or equivalent? 75MB won't go through e-mail...

BTW, holly crap, the dat file is 2.5GB?!?!?!?  No wonder 16GB's not enough Smiley Happy

Re: No sufficient memory

Builder
Builder
Hah! I know it sounds insane to have such a file size. I feel I will miss studying mesh convergence as I will be stopped by the memory restriction (using the mesh that the PC can carry it is analysis/size).

Please find the model files from here

https://www.dropbox.com/s/t7yrod0ft3cv451/Model1_i.zip?dl=0&m=
Thanks

Re: No sufficient memory

Legend
Legend

Wazy wrote:
Hah! I know it sounds insane to have such a file size. I feel I will miss studying mesh convergence as I will be stopped by the memory restriction (using the mesh that the PC can carry it is analysis/size).

Please find the model files from here

https://www.dropbox.com/s/t7yrod0ft3cv451/Model1_i.zip?dl=0&m=
Thanks


One of these silly models, uh?  I opened the NX file, it's a block with an enormous number of elements, there's really no way around it but have more memory on your computer. I thought it was an intricate model that was causing issues with contact or other type of complex stuff, but it's just an ungodly number of brick elements.  I can see why the glue joint creation has a hard time, there are 100,000s faces to deal with...

 

If you only have 16GB of RAM, then it's going to be misery all around (tell me it's not a laptop!). This is going to create an enormous op2 that will be challenging to deal with.

 

You might be able to run with the sparse solver if you create a large swap space, but you're going to be running in swap, so ouch Smiley Happy 

 

Honestly, with 20% or even 10% of the element count, you should already have some pretty good mesh convergence, don't kill yourself Smiley Happy

Re: No sufficient memory

Builder
Builder
Ram memory is 32GB, and it is a workstation.
For glueing here the purpose is to constrain components together. the best practice to simulate was using Tie constrain in Abaqus. Is there an alternative here to avoid the glueing. Does constrain components in NX to do the same of tieing of Abaqus? I am not sure!
You are correct,
I am working on reducing this number of elements to the reasonable one. some limitation (tiny curves) that affect the overall mesh fining and number!
If you have any suggestion for meshing will be a great help

Thanks

Re: No sufficient memory

Legend
Legend

Wazy wrote:
Ram memory is 32GB, and it is a workstation.
For glueing here the purpose is to constrain components together. the best practice to simulate was using Tie constrain in Abaqus. Is there an alternative here to avoid the glueing. Does constrain components in NX to do the same of tieing of Abaqus? I am not sure!
You are correct,
I am working on reducing this number of elements to the reasonable one. some limitation (tiny curves) that affect the overall mesh fining and number!
If you have any suggestion for meshing will be a great help

Thanks


Glues would be like ties, yes.  Or you could use RBE3 but that would be a lot of them too...  Don't you want to use mesh mating conditions and have the nodes merged instead of ties?  Maybe not, I'm sure you have your reasons.

 

BTW, since you mentionned Abaqus, the Abaqus people like to throw big words like "hourglassing" and "shear locking" to sound smarter.

 

Nastran users respond "LOL" to this: Nastran's brick element formulation is far more sophisticated and takes care of the majority of these issues with bubble functions (sounds way cooler than "hourglassing" doesn't it?).  Bottom line is this: when Abaqus needs 3-4 elements, Nastran performs excellently with only 2. 

 

That probably means you can already remove 1/3 of your elements simply because Nastran fights hourglassing & shear locking with bubbles Smiley Happy