Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
CQUAD4 in bending test
#11
A few notes:

- If all the plies are 0, then why go through the CLT process? Just use the beam bending formula as you did for the isotropic and the modulus that runs along the 1.5m direction. In other words, I was expecting some simple hand calcs like before.

- For a more general composite, I don't think you want the D matrix terms in this case. Rather, you should use the d matrix values. The D term would be used in a scenario where anticlastic curvature is resisted (I attached a snip from my book).

- I didn't follow the python code all the way through, but remember D (and d) are "per width" stiffnesses so you need to adjust them by the width. This will flush out if you do the simple PL^3/(3EI) calculation as well.


Attached Files Thumbnail(s)
   
Reply
#12
(08-28-2020, 08:16 PM)Admin Wrote: A few notes:

- If all the plies are 0, then why go through the CLT process? Just use the beam bending formula as you did for the isotropic and the modulus that runs along the 1.5m direction. In other words, I was expecting some simple hand calcs like before.

- For a more general composite, I don't think you want the D matrix terms in this case. Rather, you should use the d matrix values. The D term would be used in a scenario where anticlastic curvature is resisted (I attached a snip from my book).

- I didn't follow the python code all the way through, but remember D (and d) are "per width" stiffnesses so you need to adjust them by the width. This will flush out if you do the simple PL^3/(3EI) calculation as well.

Thanks - just to clarify:

1) You are right, the current example is trivial, but CLT is a general method that could be used for any laminate.
The simple hand calc results in deflections very close to CLT : 
# Analytical (Bernoulli beam):
I = (1.25e-04*12)**3*0.15/12 
F = 0.1
L = 1.5
E = 9.0e9
Zmax  = F*L**3/(3*E*I) 
Zmax = 0.296296

Now we have 3 results: 
Mystran =0.03613559 m
CLT = 0.296250 m
Beam =  0.296296 m


2) Could you share the reference please? I'm not familiar with a d matrix. 

3) adjustment for width is in the code, where the bending curvature of the beam is calculated from Ky = D^(-1) * [0, M/width,0]
Reply
#13
I have fixed the problem causing a crash with plate pressure load. Attached is a QUAD4 problem that you asked about. It is a thick plate and the bdf file is attached (you will have to change the extension as the forum would not allow .dat or .bdf). The theoretical thick plate solution for the max displacement is:

Max displacement: theoretical = 4.294E-07


Attached Files
.txt   AQ3U2S010-MIN4T.txt (Size: 16.17 KB / Downloads: 5)
Reply
#14
O_Stodieck, the book is "Practical Analysis of Aircraft Composites". The [d] properties are just from the [a], [b], [d] instead of the [A], [B], [D]. You can also use Exb (or Eby depending on the orientation) for this problem as well. All of those calculated in eLaminate here btw: https://www.espcomposites.com/software/download.html

Regarding the issue, we know that the isotropic is OK (so its probably not an element issue or general issue). I would first try just a MAT8 orthotropic material (no PCOMP) and make sure that is behaving properly. If so, then we can look that the PCOMP further. The fact that you said changing the ply angles had no effect may mean there is an issue. But first I would like to isolate the issue to either the MAT8 side (as a general orthotropic material, but not a composite) or the PCOMP side. Then I can take a deeper look into as well...but something seems odd. I have not used composites in MYSTRAN, but that is something we need to make sure is correct.
Reply
#15
(08-29-2020, 12:59 AM)drbillc Wrote: I have fixed the problem causing a crash with plate pressure load. Attached is a QUAD4 problem that you asked about. It is a thick plate and the bdf file is attached (you will have to change the extension as the forum would not allow .dat or .bdf). The theoretical thick plate solution for the max displacement is:

Max displacement: theoretical = 4.294E-07
Thanks Bill. I just ran the file and am getting a max deflection of 4.280877E-07, that seems fairly close to the theoretical, so no problems there.

(08-29-2020, 07:57 PM)Admin Wrote: O_Stodieck, the book is "Practical Analysis of Aircraft Composites". The [d] properties are just from the [a], [b], [d] instead of the [A], [B], [D]. You can also use Exb (or Eby depending on the orientation) for this problem as well. All of those calculated in eLaminate here btw: https://www.espcomposites.com/software/download.html

Regarding the issue, we know that the isotropic is OK (so its probably not an element issue or general issue). I would first try just a MAT8 orthotropic material (no PCOMP) and make sure that is behaving properly. If so, then we can look that the PCOMP further. The fact that you said changing the ply angles had no effect may mean there is an issue. But first I would like to isolate the issue to either the MAT8 side (as a general orthotropic material, but not a composite) or the PCOMP side. Then I can take a deeper look into as well...but something seems odd. I have not used composites in MYSTRAN, but that is something we need to make sure is correct.

Thanks for the references. 

I tried 2 more approaches:
  1. replacing the PCOMP with a PSHELL as suggested - the results are the same, i.e.  T3 = 3.613534E-02 (although a lot more information messages are generated, related to shear stiffness)
  2. replacing the MAT8 with a MAT2, where the G matrix terms are calculated from [D]*12/T**3 - again the results are the same with T3 = 3.613380E-02.
So it looks like the error is not limited to MAT8 or PCOMP... (see input files attached)


Attached Files
.txt   shell_bending_Pshell_MAT2.txt (Size: 121.74 KB / Downloads: 4)
.txt   shell_bending_Pshell.txt (Size: 121.67 KB / Downloads: 3)
Reply
#16
If you then change it to isotropic, does it work? What I am trying to determine is if the issue is related to orthotropy in general. It seems that might be the case.

I think the best thing to do would be create the simplest deck that demonstrates the issue. For example, I would not use any load scaling factors and I would use round numbers for the material properties, etc. Also, it will be easier if you don't use up all the entire space in the field (better separation of the numbers that includes spaces). This will be possible if you use whole numbers and it will be easier to read the deck for those that don't have column editors or less familiar with this format. Its still at a point where we have a better idea what is going, but I would encourage you to make the deck as simple, basic, and readable as possible to expedite resolution.
Reply
#17
(08-30-2020, 04:45 PM)Admin Wrote: If you then change it to isotropic, does it work? What I am trying to determine is if the issue is related to orthotropy in general. It seems that might be the case.

I think the best thing to do would be create the simplest deck that demonstrates the issue. For example, I would not use any load scaling factors and I would use round numbers for the material properties, etc. Also, it will be easier if you don't use up all the entire space in the field (better separation of the numbers that includes spaces). This will be possible if you use whole numbers and it will be easier to read the deck for those that don't have column editors or less familiar with this format. Its still at a point where we have a better idea what is going, but I would encourage you to make the deck as simple, basic, and readable as possible to expedite resolution.

Yes - an 'isotropic' MAT8 property works. 

I simplified the model and am attaching 3 versions, which are identical except for the MAT8 properties, which should outline the problem as clearly as we can get it right now. (Note, due to rounding and less elements, the model is not longer the same in geometry and properties as in previous posts). 

The _ref model has E1 = 1.3E11 , E2 = 9.0E9
The _soft model has E1 = 9.0E9 , E2 = 9.0E9
The _stiff model has E1 = 1.3E11, E2 = 1.3E11

The table below compares the analytical beam deflection at the tip to the Mystran prediction. 
There is a good match for the _soft and _stiff models. 
The _ref model is 87% out and matches neither of the isotropic models (in theory it should match the _soft model).


Attached Files Thumbnail(s)
   

.txt   mat8_ref.txt (Size: 2.78 KB / Downloads: 8)
.txt   mat8_soft.txt (Size: 2.78 KB / Downloads: 7)
.txt   mat8_stiff.txt (Size: 2.78 KB / Downloads: 4)
Reply
#18
Thanks, this looks easier to read and work with. I will have a look when I get a chance and talk with Bill about it. We should be able to get to the bottom of it. We should probably investigate beam and solid elements as well (the issue may be common).
Reply
#19
mat_ref


Attached Files Thumbnail(s)
   
Reply
#20
I look into it further and the MYSTRAN output does not seem correct. I talked with Bill about it and he is going to look into it.
Reply


Forum Jump:


Users browsing this thread: 3 Guest(s)