When you have 2 bodies bonded together with a sharp corner, a discontinuity exists and the stresses computed near the discontinuity are meaningless. My question is, are the free body forces also meaningless near the discontinuity?

Thanks.

When you have 2 bodies bonded together with a sharp corner, a discontinuity exists and the stresses computed near the discontinuity are meaningless. My question is, are the free body forces also meaningless near the discontinuity?

Thanks.

I'm attempting to determine weld size using the procedure contained in the attachment "116_Welds.pdf" which references "Determination of Weld Loads & Throat.pdf". I am working on a tank with under-head legs. The legs are W6x15, the lower tank head is 2:1 elliptical shape. Everything is modelled as surface bodies in a single part file.

The bottom of the legs are fixed. I applied a remote load (rigid connection) to the top edge of the head. The location of the remote load is at the tank's centroid. The load is primarily a vertical load with a much smaller horizontal load.

I used split faces on the legs to space the nodes 1/4" apart with an instance of 1/8" node spacing on the end of the beam's flanges. I also used split faces on the head as can be seen above. I used draft quality mesh, with global bonded compatible mesh settings, but the mesh didn't seem to be compatible. I checked all my settings and everything seemed ok, but clearly the mesh is off.

So I added more split lines to the head even though I don't think its required.

This mesh satisfied me, so I ran the study.

To be continued...

Here are the free body forces on the first three nodes of one of the beam flanges:

The forces drop significantly over a very small distance. The first and second nodes are 0.125" apart in the z-direction. The second and third nodes are 0.25" apart in the z-direction.

To finish out the rest of the process you convert the load components into a resultant force per inch. Node 1 = 5667 lbs/in; Node 2 = 475 lbs/in; and Node 3 = 211 lbs/in.

And that's when I became concerned. That just doesn't seem realistic.

I'm currently trying to recreate the example problem that I was following. I'll post up what I find.

I agree, but the free body forces still seem off. I think the end node is way higher than it should be for how fast the forces drop off. Look at the resultant force per inch that I posted yesterday. What it boils down using the procedure is the 1st node requires an 11/16" nom. weld, and the 2nd and 3rd node only require a 1/16" nom. weld.

What happens if you continue past the 2nd and 3rd nodes? If you've got a trend line for every node along that interface, and it's just the last point that suddenly spikes, perhaps you can extrapolate for the last point then multiply by the stress concentration coefficient to get a more realistic result, given that the end point of that connection will be a point of stress concentration in the real world.

Totally forgot about this thread, to which I want to make a correction to my previous statement.

Whether or not stress singularities will effect your internal force results depends on how the software calculates said forces. If the software is using the stiffness matrix and the displacement field solution to calculate the internal forces, then no, stress singularities will not effect the internal force solution. However, if the software is integrating the stress over the area to calculate the internal forces, then stress singularities will effect your internal force values. I did some digging around and I cannot find out which method SW uses, so you might want to do some testing with a very basic model to gain some insight.

No; typically the reaction loads will not be impacted by stress singularties. However, note that convergence on energy norms does not imply convergence on internal forces. If you primary concern are internal loads, then make sure that they converge.