I have kind of big model & running a sol 106. After hour stop with error and no so much details.
It should be related to the memory but not sure how to fix it
, here are some details
I allocate 7.5GB and changed the scratch directory to a big destination
This was not enough to solve the issue
please see the attached files
The iterative solver failed, better use the DIRECT SPARSE solver, is more robust.
Yes, the default solver with NX NASTRAN is the DIRECT SPARSE solver:
In summary, to speed the solution time and increase the NX NASTRAN performance (running with any pre&postprocessor, say SIMCENTER or FEMAP) you can do the following:
You can learn more about NX NASTRAN PERFORMANCE visiting my blog:
Thanks Blas for the very valuable details here.
So how to define the buffsize? is it inputted in the .dat file? (I assume that before running the software write the input file then edit the dat file and insert the command in the first line. afterwards, we can run the software?
is not there a way to define the buffsize it in the Entry bulk parameters?
The temp folder destination is it changed through the environmental variable API_DEFAULT_DIR, please?
You need to locate the nast11.rcf file, this is the nastran resources configuration file, Go to "../nastran/conf" and edit the file nast11.rcf using NOTEPAD, there make sure to edit the nastran keyword SDIR to point to the correct location of SCRATH directory. For instance, in my system SDIR=C:\SCRATCH means that the nastran scratch will be created in the location C:\SCRATCH, where I have a fast SSD drive, plenty of space.
For superior nastran performance, is better to have a dedicated separate fast SSD drive for only nastran scratch, like D:\SCRATCH using a fast PCIe m.2 SSD drive with say 1 GB of disc space!!.
Don´t copy, the highlighted ones are OK.
You can edit the nasxx.rcf file any time you like, but to take effect the changes you need to execute again the NX NASTRAN solver.
Unfortunately, no change at all. I am not sure if there are more things to do. I am using less than my original model with the least order element and it is run with one step/2hrs. Actually, I'm pretty sure now it is a little bit slower than earlier case17hr/8 vs 15hr/8steps) I am not sure if I used the original model how much running time would cost?
The challenge we needing several runs before validating the model.
Any suggestion, please?