ABAQUS/Explicit: Introducing imperfection in truss like structure

[ShadowWarrior]

How do I introduce geometric imperfections in the form of slight deflections in the transverse direction to the truss axis? -

http://i.imgur.com/LxUsMgF.jpg

I tried the conventional method of doing Buckling analysis and extract first 10 Eigen-modes. Later, using *IMPERFECTION keyword in the final analysis with the 10 Eigen-modes scaled to introduce surface imperfection. But the resultant imperfection is not in the transverse direction to the truss axis.

 

[ShadowWarrior]

Anyone?

 

[cooken]

What boundary conditions did you use for the modal analysis? If you leave only the transverse DOFs free, then the mode shapes should be only in that direction.
Alternatively if you leave the system free - if the transverse stiffness is much larger than other directions, the first 10 modes may not include transverse deflections so you may need to extract more, then with *IMP only select the transverse modes and scale them accordingly.

If you are like me and want ultimate control, you can also write your own script to introduce a custom imperfection. You can do this outside of Abaqus, using initial node coordinates to generate a custom input file and again use the *IMP card, or you can use a python script (or do it by hand even if you have very few nodes) directly in CAE using the mesh edit method - editNode()

 

[ShadowWarrior]

@cooken - Thanks for your reply.

Boundary Condition:
Bottom nodes - Fixed in all direction
Top nodes - Fixed in lateral direction, free in vertical direction, 1N downward force.
How can I leave only the transverse DOFs free, when the 3D XYZ axes make 45 degree angle with the struts?

I have 400K elements, so doing in hand is not possible. Can you direct me to an example .py script to introduce custom imperfection?

 

[cooken]

You can create a user-defined coordinate system with one axis aligned to the truss axis, and another aligned with the transverse direction. Then create the BCs through that CSYS - Abaqus will automatically take care of the coordinate transformations for you.

Example python block:
truss=mdb.models[''].parts['']...
for n in range(len(truss.nodes)):

x = truss.nodes[n].coordinates[#]/length_truss #Axial coordinate along truss normalized to 1.​

offset = f(x) #Some algorithm (sinusoidal/logarithmic etc..) as a function of the axial location along the truss.​

part.editNode(​

nodes=truss.nodes[n:n+1], #n+1 because it expects a sequence​

offset1,2,3=offset[], #Choose whatever number represents the transverse direction​

projectToGeometry=OFF)​

You can make the above as complicated as you want. Try it on a simple case first (eg. truss aligned with a global axis) to make sure your algorithm is working. Please read the scripting reference documentation 18.2.8 (2016). Also note this will not be particularly fast, so if that's an issue you can generate text file containing the offsets and use that with *IMP

 

[ShadowWarrior]

Its actually seems to be out of my league, provided the little knowledge I have in scripting. Anyway to do something similar in CAE?

 

[cooken]

This really isn't difficult, I highly recommend you do some light reading such that you change leagues. Using the scripting interface is quite powerful and does enable you to do things that just wouldn't be practical otherwise.

You can still use the other methods I suggested above, define a new CSYS to apply boundary conditions while leaving transverse DOFs free, or just extract higher modes until you see the ones you want and then use those in your *IMP.

 

[ShadowWarrior]

Extract higher modes means requesting 20 or 30 modes? Currently I am requesting 10.

 

[cooken]

Yes.
I assume you are needing the transverse shapes because that is one the expected imperfection orientations? But then if you don't see transverse displacements until much higher modes then how likely is it to buckle like that?

Also just looking at your original post...the "resultant" imperfection is a superposition of each mode shape right? Have you actually checked to make sure none of those 10 are transverse?

Can you upload your CAE file?