I hit Run Simulation on my electronic box FEM on Thursday, and it is still running. I am running linear modal frequency analysis on it.
I don't have any errors or anything, doesn't seem to have crashed. Is this normal? It's a pretty complex assembly but should it be taking this long?
I feel safe in saying that 12 days is abnormal. Have you used the task manager (ctrl+shift+esc) to check that the analysis is running? (The "Process" will be ssap something or srun something, such as "ssap1.exe").
But the runtime depends on a number of variables, such as
- How many frequencies are you solving for? (If you did not change the default, then the answer is 5.)
- The model size is determined by the number of nodes in the model (or more accurately by the number of equations or number of DOF). One of these three should be near the beginning of the log/statistics file (ds.lgm) or summary file (ds.ml), both of which are located in the folder named "whatever your model name is.ds_data\design scenario number". What size is the model?
- What version of the software are you using? (This is shown in one of the first 3 lines of the log or summary file.)
- Did you change the solver from the default to a different type? If so, which one? (Hopefully not the slowest solver which is the subspace solver.)
16 years experience with Simulation Mechanical
Thank you John for your response. To answer your questions, here is the info from the checklog:
Autodesk (R) Simulation Natural Frequency (Modal)
Version 2014.01.00.0025-W64/X64 26-Jul-2013
Copyright (c) 2013, Autodesk, Inc. All rights reserved.
DATE: OCTOBER 31, 2013
TIME: 04:16 PM
INPUT MODEL: C:\Users\user\Documents\LCRD\LCRD FEM\812616TowerB_2013-10-29FULLMODEL.ds_data\1\ds
PROGRAM VERSION: 201401000025
alg-win-x64.dll VERSION: 201401000025
agsdb_ar-win-x64.dll VERSION: 201401000025
algconfig-win-x64.dll VERSION: 201401000025
solvercallback-win-x64.dll VERSION: 201401000025
amgsolve.exe VERSION: 360000000
**** Dynamic linear modal analysis
**** Nodes | Parts | Loads | Freqs | Mode
1293136| 113| 1| 5| 1
Options executed are:
**** OPENING TEMPORARY FILES
NDYN = 1
**** BEGIN NODAL DATA INPUT
**** END NODAL DATA INPUT
And here is the data from the ds.lgm
1**** CONTROL INFORMATION
number of node points (NUMNP) = 2638314
number of element types (NELTYP) = 136
number of load cases (LL) = 1
number of frequencies (NF) = 5
analysis type code (NDYN) = 1
equations per block (KEQB) = 0
bandwidth minimization flag (MINBND) = -1
gravitational constant (GRAV) = 3.8640E+02
number of equations (NEQ) = 7943832
**** PRINT OF NODAL DATA SUPPRESSED
I am solving for the first 5 frequencies.
I don't remember changing the solver so hopefully it is just default.
If the model is too big should I be looking at it in sections? What can I do to make it more manageable?
Thanks so much!
Is there a limit for number of nodes? Any rule of thumb there I can follow?
Is there a simple way to split up my model and analyze it in parts?
What is the best approach here?
I can relate to your issue on time for an analysis to run. Usually, the analyses run fairly quickly, if all is well and they run for days if there is an issue with your model. I usually run large multi part models of static analyses. My method is to use simple modeling then get increasingly more realistic. An example is to use default contact initially, then on subsequent runs change to surface contact to add reality. This will take significant more time to run with the surface contacts, but you will know that there are fewer other modeling errors with the previous simpler runs.
There are so many things that can affect the run time that it is very hard to point out the ONE that is affecting your model.