Weld Capacity Analysis Using FEA Results

[Ray2020]

When working on a simple beam model the forces extracted from joints in that model show forces correlating to an acceptable weld capacity.
When the same structure is modelled with finite elements (plate/shell elemets), sometimes you might see that for this same joint a significant portion of the length displays forces correlating to a non acceptable weld capacity. This is because FEA can provide better resolution. If the average force per unit length of the entire length was taken then this value will correlate to an acceptable weld capacity.

What do you do in these circumstances? Do you just take the average along the entire length?
Not everyone performs FEA on structures, if someone is using a simple beam model in combination with standard AS 4100 to perform their calculations then they wouldn't notice this problem.
Or do you just keep assuming a thicker weld until results correlate to a passing capacity?

 

[SWComposites]

Probably depends on whether you can assume the weld yields before failure or not.
Probably also depends on whether you believe the FEM generated distribution is more accurate. I would not automatically assume the FEM distribution is “better”.

 

[la belle vie]

Probably not. In real life the weld will yield at the stress concentrations and redistribute.

 

[human909]

Not everyone performs FEA on structures, if someone is using a simple beam model in combination with standard AS 4100 to perform their calculations then they wouldn't notice this problem.

Possibly the problem is with you using simple FEA on structures? Rather than suitable analysis in line with AS 4100.

As the others above have suggested it seems like your FEA analysis isn't accounting for plastic behaviour of the material.

FEA is a great tool but it needs to be applied suitably and interpreted suitably.

 

[MotorCity]

For a simple beam model, its hand calcs using first principles every time. The issue with using FEA for something like this is that your FEA results give very detailed and precise results, but your model (or any model for that matter) is based on gross assumptions on the loading, fixity, etc. For example, if you have a uniform load of 10 psf, is the load really uniform? does it really cover the entire area? or is it more like 9.56 psf? Same goes for fixity and support conditions....The point is that the degree of accuracy of your output (FEA results) needs to match the degree of accuracy of the input (loads, fixity, etc), otherwise there will always be a huge disparity between FEA results compared to your simple beam model results. To answer your question, I take the force/weld length to determine the required weld strength. FEA is basically showing you that the force along the weld is not uniform and there are higher force concentrations along its length, which is sort of like a shear lag effect.

 

[ThomasH]

Ray2020

I often model beams using plate/shell elements. But I would not say that is the best approach if you actually want to include the welds in the model.

I don't know the AS 4100 but in the Eurocode there is a limit for the length you may include in the calculation. If you make an analysis of a long weld you can see that the stress in the weld varies along the length. The same will happen if you put enough bolts in a row.

If you actually want to include the weld in the model, I would probably use volume elements. I had a colleague who did a PhD on fatigue and that was the approach in that case. I also know that it was used in projects but that was not the normal method.

 

[Ray2020]

Possibly the problem is with you using simple FEA on structures? Rather than suitable analysis in line with AS 4100.

No.

 

[bigester93]

what yield stress of weld do you consider?

 

[human909]

No.

Well since you responded quite bluntly I'll respond in kind. When I used the word "possibly" I was being polite. What I really meant is that your use FEA is the problem. Either your modelling technique or more likely your interpretation of the results. You are observing a problem that is almost certainly not there except in your flawed analysis. Structural codes are based on decades of built knowledge usually based on theory and experimentation.

I say this as somebody highly experience with AS4100 and FEA on steel structures. That alone doesn't give me greater authority. But given you are a mechanical engineer with only a few years experience who is asking this question, I would think it would be wise to listen to some of the answers given.

 

[Ray2020]

Probably depends on whether you can assume the weld yields before failure or not.

In real life the weld will yield at the stress concentrations and redistribute.

what yield stress of weld do you consider?

The fillet weld strength limit state according to AS 4100 is determined by an equation where the tensile strength is used along side an inbuilt 0.6 factor in the equation and a 0.8 capacity factor for the fillet weld. Even though the equation is considering tensile strength, it is factored to a degree that the acceptable criteria is lower than the yield stress of a material any way.

 

[Ray2020]

I take the force/weld length to determine the required weld strength. FEA is basically showing you that the force along the weld is not uniform and there are higher force concentrations along its length, which is sort of like a shear lag effect.

Thanks for the response MotorCity. Yes, that's what is happening.

 

[Ray2020]

Well since you responded quite bluntly I'll respond in kind. When I used the word "possibly" I was being polite. What I really meant is that your use FEA is the problem. Either your modelling technique or more likely your interpretation of the results. You are observing a problem that is almost certainly not there except in your flawed analysis. Structural codes are based on decades of built knowledge usually based on theory and experimentation.

No.

 

[ThomasH]

The fillet weld strength limit state according to AS 4100 is determined by an equation where the tensile strength is used along side an inbuilt 0.6 factor in the equation and a 0.8 capacity factor for the fillet weld. Even though the equation is considering tensile strength, it is factored to a degree that the acceptable criteria is lower than the yield stress of a material any way.

Regarding this I have some comments. Sinece I don't know AS 4100 but I do know a bit about other codes and mechanics. I assume that this concerns a long weld and it probably supports the load in shear. That probably means that 0.6 is short for 1/sqrt(3) = 0.577 and that is von Mises yield when applying shear instead of tension. As for the 0.8, that may be a safety factor. I don't know how the safety principles are set up in that code.

If I use an equation from the Eurocode:

fu is ultimate strength
sqrt(3) is von Mises, tension to shear stress.
beta_w is a factor consitering the base material (beam material). For S355 steel it is 0.9.
gamma M2 is a safety factor that is selected in the national annexes, usually ~1.2 since the equation is based om ultimate strengt.
The base for safety in the Eurocode is factors on the load.

You said in your first post that "not everyone performs FEA on structures". In my experience many engineers do use FEA and the most common element type is beam elements, but it is still FEA . For the type of problem you work with I think plate/shell elements could suffice but you need to use non-linear analysis. The effects you are dealing with includes deformations and that requires non-linear methods.

If you describe your analysis better we may be able to help you. My impression is that there is something wrong in the analysis, not a flawed code.