Different results between NASTRAN and ABAQUS

[dankoonFANTOMAS]

Hello everyone,

I made the same analysis in NASTRAN and ABAQUS in order to compare their performance and results.

It's a modal dynamic analysis (half sine shock), the shock is 0.011 s long and my analysis time is 0.22 s in order to see the behavior of the structure after the shock.

In ABAQUS I used the following command lines:

*STEP, INC=200
*MODAL DYNAMIC, CONTINUE=NO
0.0011 , 0.22
*MODAL DAMPING,MODAL=DIRECT, DEFINITION=FREQUENCY RANGE
0.0,0.03
400.0,0.03
*SELECT EIGENMODES, DEFINITION=FREQUENCY RANGE
0.0,400.0
*BASE MOTION,DOF=1,AMPLITUDE=CHOC_11ms,SCALE=9.81,TYPE=ACCELERATION

Where CHOC is an input file where my half sine coordinate are set.

For NASTRAN I used the following commands:

SOL 112
DLOAD = 101
TSTEP 100 200 0.0011 1
TLOAD2 101 101 ACCE 0.0 1.1E-2 45.45455270.0
TABDMP1 4 Q
+ 0.0 16.7 400. 16.7 ENDT
SPCD 101 174482 1 9.81
ENDDATA

SOL 112 is based on a cosine so I used a phase of 270 in order to obtain a sine.

The thing is that if I plotted the acceleration OUTPUT of the same node from both analysis I have a similar behavior but I observe differences (see attached picture).

I do not know where the differences come from, do you have an idea?

The blue curve is NASTRAN and the orange one is ABAQUS

Thks

 

[btrueblood]

You have a very slight differnce in fundamental freq. There may be a difference in how the finite elements are treated, differences in meshing, or differences in how the diff. eqtn. is solved (newtonian iteration or runge-kutta or ...?)

 

[dankoonFANTOMAS]

The analysis is based on the same FEM model (made in Hypermesh and exported in .inp for ABAQUS and .dat for NASTRAN) so I have the same number of nodes, elements, materials.

I made a previous modal analysis and I obtained the same frequencies, modes and effective masses with both codes.

Yes this is MSC NASTRAN.

I first thought that it was problem of a difference between the sine defined by ABAQUS and NASTRAN but I plotted the curves of total acceleration at the excitation nodes to see if both codes used the same input excitation. They are the same, so I think maybe the problem comes from the damping coefficient.

n NASTRAN I have the following values for the damping (I used CRIT type): Q=30, G/2=1/2Q so G=0.03333 and CRIT=0.01667

In ABAQUS I put G as the modal damping values (don't know if it's the right approach)

 

[btrueblood]

"maybe"? Of course the damping affects the fundamental frequency.

 

[dankoonFANTOMAS]

Of course I agree it is important in this kind of analysis my point was about the origin of the difference between results.

Well in fact I am not sure it is the source of my problem here. I just run a sine sweep frequency on both codes and I obtained the same curve for the acceleration vs frequency (following the direction of the sine sweep). So I suppose my damping coeff are good here.

So I join you on your first impression it's maybe the way the diff eqt are solved.

 

[rb1957]

"a cosine so I used a phase of 270 in order to obtain a sine" ... 270deg ? not 90 deg ??



another day in paradise, or is paradise one day closer ?

 

[dankoonFANTOMAS]

In order to have a positive input signal.

I agree with you, I first put 90 deg but my input signal was negative and I had to apply a -1 factor to my shock in order to have the same input than the one in ABAQUS. I then tried 270 deg and I have a positive signal without adding my -1 factor.

By the way in both case (phase=90 and -1 factor and phase=270 without -1 factor) give the same results (the curve in my first post)

 

[rb1957]

"half sine shock" ... =0 at t=0, =max at t=0.011 ? (a sine wave period = 0.044sec ?)

sin(x) = cos(x+270)

another day in paradise, or is paradise one day closer ?

 

[dankoonFANTOMAS]

Yes

So adding a phase of 270 is right we are agree?

My half sine shock is 0.011s long so the max is at 0.0055s, so a period of 0.022s.

 

[rb1957]

sounds right ...

another day in paradise, or is paradise one day closer ?

 

[nlgyro]

The critical damping you calculate is 0.01667, but that not what you have entered in your nastran SOL112??

 

[nlgyro]

just eyeballing your 2 inputs....

You use a 0.03 (or 3%) modal damping in ABAQUS analysis whereas for NASTRAN you use a Q damping of 16.7. Which equates to a critical damping ratio of 1/(2*16.7) = 3%, so your modal damping becomes =2*3% = 6%. This doesn't explain the phase shift that you see but maybe you can redo your analysis with consistent inputs for both solvers.