WHy the beam elements in LS-Dyna behave so stranngly?!!!

Zack Vizeman

unread,

Dec 4, 2020, 7:58:41 AM12/4/20

to LS-DYNA2

Dear all,

I have analyzed a simple two-member frame-like structure, right out of most FE textbooks, with the first member sloped at 60 deg (with respect to the horizontal axis) and the other member running horizontally. The first member is fully restrained at its bottom end and the right end of the horizontal member is also restrained fully at its right end. Ther is a vertical concentrated load applied on the mid-height of sloped member and the top horizontal member is subject to a UDL.

I analyzed the frame with another FE software using only two elements and got the exact values of bending moments on the nodes. I also used 4 elements to model each member and got the same moment values on the nodes.

I then used (ELFORM) 4 and 13 (Timoshenko) beam elements of LS-Dyna, and got large discrepancies in moment values (obviously and naturally more discrepancies on the nodes on beams under UDL. I then started increasing the number of elements in modeling each member of the fame (from one for each member to 40!). The results improved at the 40 element mark but still not very close to the theoretical values.

Could any of the LS-Dyna guru tell me what is going on with such a simple FE analysis using two of supposedly most accurate beam elements in NISA?

Regards,

Zack

Daniele Speziani

unread,

Dec 4, 2020, 8:47:22 AM12/4/20

to LS-DYNA2

Hi Zack,

What you reported is strange, I am adding a paper that could help you.

If you are still in trouble can you share your model?

Best Regards from Italy

Daniele

LS-DYNA_ULM_H-I-31.pdf

l...@schwer.net

unread,

Dec 4, 2020, 10:01:21 AM12/4/20

to Zack Vizeman, LS-DYNA2

I think what you are observing is the lack of “work equivalent moments” in your model, and how different software include or ignore this work equivalency.

Consider your horizontal beam with the uniform distributed load. The LS-DYNA internal representation for this load is a point load at end of the beam.

Obviously, under this loading the beam will never deform and hence develop end moments. LS-DYNA depends on the user to add to the end point loads the work equivalent end moments that will cause the beam to bend as expected. Other software might opt to add this additional loading internally.

Note: As you have observed, the magnitude of the work equivalent moments decreases as the beam length is refined. The intermediate element internal work equivalent moments cancel and the model is left with only work equivalent end moments dependent on the length of the beam element nearest the ends.

--len

--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/61e3f5c8-55f3-4cbb-b0ba-ce983fce7925n%40googlegroups.com.

Zack Vizeman

unread,

Dec 4, 2020, 10:59:20 AM12/4/20

to LS-DYNA2

Dear Len,

Thank you very much for the quick response. However, I believe the explanation defies everything I have learned in my FE courses!

You see, I just used the simple Euler-Bernoulli’s beam element stiffness matrix in Excel and got the exact bending moment values at the nodes. Of course, I converted the UDL to equivalent nodal loadings (WL/2 and WL^2/12 for reaction force and moments, respectively).

I also changed the UDL to equivalent concentrated force and applied it on the mid-span of the beam. Still the results produced by LS-Dyna are way off.

Now, the question is simple – how come one of the least expensive FE software could produce almost exact results with only two elements, while one of the most sophisticated FE software in the world cannot produce close results even using 20 elements?!

Regards,

Zack