cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

welded assembly questions

Highlighted
Newbie

welded assembly questions

Dear experts;

I have a 3D trussed structure made of square tubing supporting a thin plate
all modeled as Shells which I am loading perpendicular to the sheet.
Two basic problems are still dogging me: 1) the membrane stiffness problem
(sheet initially loaded in bending, quickly transitioning to tension not
being exhibited), and 2) getting a realistic idea of the stresses near the
geometric discontinuities of the tube joints without modeling actual weld
fillets. Does anyone have any techniques that could address these problems?

Alan Carlson
Discern Engineering
1681 E. Hennepin Ave. Suite 180
Minneapolis, MN 55414-2562
(612)617-9258
(612)378-5503 FAX
(612)270-3926 cell
---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.677 / Virus Database: 439 - Release Date: 5/4/2004
5 REPLIES 5

Re: welded assembly questions

Alan,

Not sure about 1), I would have to see the problem. On 2), you sacrifice the ability to model things like fillets and welds when you decide to go with shells. It is a universal trade-off. When you gain something, you must give something up. You are gaining run time, but giving up 3 dimensional features when you go from solid elements to shell elements.

I realize that square tubing will fail the default solid automesher. But you can add seed points to get through that. It all depends on how bad you need to model the fillets. The solid modeling time (and run time) might not be as bad as you think. The welds can be part protrusions or separate parts in an assembly. Separate parts can have different material properties.

Finally, you can keep your shells and add welds using one of the methods mentioned above. If you do, then move the midplanes of the tubes to the outside and move the midplane of the thin plate to the bottom.

Randy Speed, Principal
Speed Consulting, LLC
2871 Howard Road
Waxahachie, TX 75165
(972) 938-0490 ph (972) 937-2319 fax
www.speedconsulting.com

Re: welded assembly questions

Thanks, Randy for the response.
concerning my question #2, I have had some advice that adding some "seed
points" along the intersecting edge where the weld would be placed would
allow me to gauge just how much of the stress was due to the geometric
non-linearity and how much was real. In other words, if the refined mesh in
the weld region gave a hot spot smaller than the weld fillet, then I could
write it off as due to geometric non-linearity. Would you agree with this
technique? and if so, do you have any tips or pitfalls that should be
avoided?

Thanks again for the help.

Alan Carlson
Discern Engineering
1681 E. Hennepin Ave. Suite 180
Minneapolis, MN 55414-2562
(612)617-9258
(612)378-5503 FAX
(612)270-3926 cell

Re: welded assembly questions

Alan,

Geometric nonlinearity is the same as large deflection, which you will most likely not have on this model. Most fea problems fall into the "small deflection" theory. Small deflection is commonly defined as when your model rotations in degrees is approximately equal to the sine of the rotation. This occurs at rotations less than about 5 degrees. Otherwise, at higher rotations you are working in large deflection theory. A diving board would be a good example. You can tell whether your part is overall deformed to a rotation of 5 degrees or more by looking at the deformed model with a 1.0 scale factor with the overlaid original.

What you may be referring to is non-linear material, which is referred to as plasticity. That is when the local stresses exceed the yield strength of the material. MECHANICA does not check for or model plasticity. In fact, there may be no "non-linearity" in this model. It would depend on how high the load is, etc..

Welds can be accurately modeled in MECHANICA using one of the techniques already mentioned, but it is very difficult to model in traditional H-element codes.

The stresses calculated in MECHANICA are not very sensitive to the number of elements when compared to h-element codes. Convergence algorithms raise the polynomial levels to achieve the final stress magnitude. The algorithms are SPA and MPA. With MPA, you can request tighter convergence (like 1-2%) in the weld elements. So although placing seed points around the welds will help the automesher, I would not intentionally add seed points until I see that the stresses are not converging.

Back to your question. I believe you are referring to geometric singularities. This is where the model geometry does not capture the actual geometry, especially where there are no fillets (or fillet material - like welds) and sharp corners exist. Also, using MPA, you can determine for sure whether the weld material is actually converging. You can also determine whether geometric singularities exist just be looking at the element mesh (or your CAD geometry that you are meshing). Singularities usually will not converge or will converge at extremely high (unreasonably high) stress levels. Theoretically, the stress in a sharp corner goes to infinity.

Your best bet, it sounds like, is to model the welds using one of the methods and make them as smooth as you can that imitate the overall shape of the actual weld. Is it concave or convex shaped? Make the weld go in between the parts to some extent by having the parts mating offset if you are going to model all solids. If you are going the make the plate and tubes shells, then this detail cannot be included.

Randy Speed, Principal
Speed Consulting, LLC
2871 Howard Road
Waxahachie, TX 75165
(972) 938-0490 ph (972) 937-2319 fax
www.speedconsulting.com

Re: welded assembly questions

Randy
you are correct, I was confusing geometric singularity with geometric
nonlinearity. In fact, my 1st question on membrane loading was really about
geometric nonlinearity.
If I understand you correctly, the seed points strategy should only be
applied if I do not see sufficient convergence in a tight convergenge MPA
analysis. What do you think of the idea of comparing the size of the "hot
spot" (uncoverged due to geometric singularity) to the size of the expected
weld as a short-cut for determining whether the weld is overloaded or not?
Because of the number of welds in this assembly, I would really like the
model to be in Shells and not create all those fillets. Do you think this
shortcut is safe? or conceptually correct?

Thanks for all the help so far,

Alan

Re: welded assembly questions

Alan,

Yes I do. If you make the weld material smooth, then there should not be any hot spots that are not real. There may be stress peaks where the shells meet the solid welds, but within the solid weld material, everything should be OK (as long as the geometry is smooth).

Good luck and happy 4th!

Randy Speed, Principal
Speed Consulting, LLC
2871 Howard Road
Waxahachie, TX 75165
(972) 938-0490 ph (972) 937-2319 fax
www.speedconsulting.com