restart with modal frequency response analysis 2

[cavva78]

Hello,
I'm using FEMAP 11.1 with NX Nastran for frequency response and random response analyses. Following tutorials I'm using the restart capability performing first a modal analysis then the frequency response (modal frequency method) using the saved database from the modal analysis. The problem is it seems to be completely useless, because in any case the freq. response analysis repeats the modal analysis and analysis time doesn't change. Anyone using FEMAP can help me about that?
thanks

 

[jbrackin]

If you are running the example problems delivered with Femap, then the analysis times will not change, the problem is too small to demonstrate any of the time benefits of restarts.
Also note that the restart logic and its performance is a function of the Nastran solver, Femap is only supplying an input that Nastran interprets and then the Nastran restart logic decides what operations to perform based on changes in the input supplied.
The way to tell if restart logic has been invoked in Nastran is by inspecting the F04 file and looking at what operations Nastran has performed. If you compare the f04 file from a restart run and the same analysis using "cold-start" run, then you can see the different path taken by Nastran.
If you have doubts about whether Nastran actually used restart logic in the run, a quick inspection of the F04 will reveal it. You should lines that look something like this for a restart run:

The following items have changed in data block GEOM4
Record :Entry :Item Record :Entry :Item Record :Entry :Item Record :Entry :Item
SPCD : :SID SPCD : :ID SPCD : :C SPCD :
9:40:50 0:01 129.0 0.0 0.2 0.0 IFPL 257 RESTART BEGN
9:40:50 0:01 130.0 1.0 0.2 0.0 IFPL 258 RESTART BEGN
9:40:50 0:01 130.0 0.0 0.2 0.0 IFPL 259 RESTART BEGN
9:40:50 0:01 130.0 0.0 0.2 0.0 IFPL 260 RESTART BEGN
The following items have changed in data block DIT
Record :Entry :Item Record :Entry :Item Record :Entry :Item Record :Entry :Item
TABDMP1 : :ID TABDMP1 : :F TABDMP1 : :G TABLED2 : :ID
TABLED2 : :X1 TABLED2 : :FLAG TABLED2 : :X TABLED2 : :Y
TABRND1 : :ID TABRND1 : :CODEX TABRND1 : :CODEY TABRND1 : :F
TABRND1 : :G : : : : : :


Hope this information helps.

Joe

 

[cavva78]

Hello Joe,
thanks for the reply. I'm not using the example problems but a model with about 100000 nodes and 800 modes in the first 2000 Hz, that's why I would like to use restart to avoid repeating the modal analysis every time I perform a random / frequency response run, it's very time consuming. Using restart seems to do nothing as the time doesn't change and I think it's repeating the modal analysis every time. I didn't look into the .f04 file, thanks for the tip I'll try to read it.

Marco

 

[jbrackin]

Marco,
I think I see what may be happening. The new RESVEC options may be causing Nastran to recalculate the modes. The default in V9 is to calculate residual vectors for response runs. If you submitted the normal modes run with no loads defined, no residual vectors are calculated. When you add the dynamic loading for the restart, Nastran sees loads are new, and the default is to calculate residual vectors, hence new modes calculations are performed. Seems like 2 options to me, turn off residual vectors in your response run( may or may not be a good idea for accuracy). You can do this in Femap on the Bulk Data Options form.
The other option option is to set up the normal modes run with loads and residual vectors activated so that they already exist for the restart.

Hopefully you have experience with residual vectors etc to be able to make correct choices. If you need help with that, then we may need more details on your problem, loading, etc to make sure the proper choices are made.

Joe

 

[jbrackin]

I just confirmed this is actually a bug in Nastran V9 which is the version that comes with Femap 11.1. Nastran is smart enough to just append the residual vectors without recalculating the modes in previous versions. So you could try running a previous version(11.01 or earlier) of Femap with V8.5 of Nastran, or wait for the MP1 release of Nastran which should be available in a month of so.


Joe

 

[cavva78]

Dear Joe,
thanks for the reply. I also tried running a frequency response analysis with restart after a modal analysis disabling the RESVEC, but nothing changed in the running time... Always stting up everything from inside FEMAP 11.1.0.

Marco

 

[jbrackin]

Marco,
There is also a chance that something else has changed in the model that is triggering the recalculation of the modes.If you can post the f04 file we might be able to decipher what has triggered the recalculation. I did run a test case where I restarted a random vib run from normal modes and turning off RESVEC did stop the recalculation, but I was careful to make sure nothing else changed.
The other thing that would be helpful(short of sending the whole model) would be to list the executive and case control section from the modes run and the restart.

 

[cavva78]

Dear Joe,
I did some other tries but I'm not sure of what happened, restart still seems to be useless to me.. I'm attaching the following files:
-.dat (without nodes and elements), .f04, .f06, .log of first modal analysis with database saving for restart (modal.zip)
-.dat (without nodes and elements), .f04, .f06, .log of random analysis with modal frequency method using restart and default resvec (modal freq restart resvec.zip)
-.dat (without nodes and elements), .f04, .f06, .log of random analysis with modal frequency method using restart and resvec,no (modal freq restart no resvec.zip)

Both random analyses require the same time and with param,resvec,no it still seems to calculate residual vectors.

I would be very glad if you help me understand!
Regards
Marco

 

[cavva78]

Sorry here are the correct links

 

[jbrackin]

Marco,
Using the current release of Femap with Nastran V9, the only way to stop the recalculation of modes is to set resvec=no in the case control section of the modes run and the random restart. If you need residual vectors for accuracy, then you will have to use Nastran 8.5(Femap 11.0) or wait until there is a Nastran maintenance release that fixes the restart logic.

You can do this inside Femap by doing the following:
1) on the Bulk Data form make sure that resvec is unchecked( this will turn off writing the param bulk data input for resvec)

2) in both the Modes run and the random restart run go to the Master Requests and Conditions form
Select "Start Text"
Type in "resvec=no"


if you use the "preview input" button in the analysis manager, then you should see the following:

DBSETDEL USROBJ,USRSOU
NASTRAN SYSTEM(442)=-1,SYSTEM(319)=1
ID ex19,Femap
SOL SEMODES
TIME 10000
CEND
resvec=no
TITLE = Normal Modes Analysis
ECHO = NONE
DISPLACEMENT(PLOT) = ALL
SPCFORCE(PLOT) = ALL
METHOD = 1
SPC = 1
BEGIN BULK


Hope this gives a usable workaround for now.

Joe