Weld Modeling in Simulation

[gnatn88]

I am trying to model a plate that is welded along the edges to a larger plate. I have tried 2 approaches separately:
1) Partitioning the face of the larger plate and applying a bonded contact between the plate edges and the edges on the surface
2) Bonding the surface of the smaller plate to the surface of the larger plate

I believe these methods are OK but the stress is not that accurate at the weld location. I need accurate stresses near the welds because I believe this is where failure is likely to occur in my model.

Can anybody suggest a better approach to modeling the welded joint?

Is it better to create parts that represent the actual fillet weld material and then bond (tie) these to the 2 plates?

Thanks!

 

[gnatn88]

After doing some research I have found this paper from 1999. Maybe others will find it useful. I would still like to hear suggestions from others on how to tackle this problem. Thanks!

Determination of Weld Loads and Throat Requirements Using Finite Element Analysis with Shell Element Models — A Comparison with Classical Analysis

http://www.aws.org/wj/supplement/Weaver/ARTICLE2.pdf

 

[321GO]

Why not hand calc's? Not applicable?

 

[KevinDeSmet]

They have added an Edge Weld connector a couple of releases ago, you will have to experiment with it results in your applications:

http://help.solidworks.com/2013/Eng...R.htm?id=381dc94158944c00a3ff8f9e2985bea2#Pg0

As far as literature, there is an old Cosmos Companion issue on the topic of analyzing welds, be mindful this was before they added the Edge Weld connector:

http://www.eng.uwo.ca/designcentre/FEA resources/116_Welds.pdf

Certified SolidWorks Professional

 

[steris]

If I understood correctly, I don't believe that method 1 will work because the contact area bonded is only edges. This would create massive stress risers.

Method 2 would produce much lower stresses than actual because the load transfer happens over the area of the smaller plate instead of only at the welds.

Assuming it's a fillet weld, could you model the weld as a separate part? You could make a triangular extrusion and then assemble it along each edge of the smaller plate? Then within Simulation, apply a bonded contact between the weld and the edge of the plate as well as the weld and the face of the larger plate.

 

[MikeHalloran]

I managed to model a complete skid in SW2010's one-body-limited demo by subtracting solids from a large prismatic block. Chamfers on some of those subtracted solids left an approximation of fillet welds. The predicted stresses were at least credible.

The stresses were also too high. If the skid was built as I was ordered to design it, the engine mount supports would have bent enough to notice under the engine's dry weight.

(
I don't know if that happened. My boss got yelled at for allowing me to waste time producing a fancy report with colored stress contours and magnified deflections. The report didn't actually cost them much. I did the whole thing in a couple of hours on a Saturday, on my time.

Not long after that, before the product was built, I got caught in a mass layoff.

No good deed goes unpunished.
)

Mike Halloran
Pembroke Pines, FL, USA