Weld stress analysis using axisymmetric model

[barrind]

I'm after some guidance on how to analyse welds based on stresses derived from an axisymmetric model. Basically the shape I am modelling is a can under internal pressrue with top and bottom covers fixed by single sided fillet welds. The designers have asked for confirmation that the weld throat size they have specifed is sufficient. To cut a long story short the bottom cover is fairly complex geometry and so although they have done some traditional hand calcs they want this backing up with FEA. Up to now I have only done fatigue assesments of welds, not actual design analysis and I am unsure how to approach this from FE when comparing the stresses to the allowable shear stress of the weld material. Would this be a use for the stress linerisation functions in Abaqus or similar? I am hoping using an axisymmetric model will make extracting the relevant stresses easier. Any advice would be of help. Thanks.

 

[ishvaaag]

With your problem I would make a 3D model, assume single material, and apply the inner pressure, and support or whatever, then would go after the VonMises stresses. Then would consider the limit for fatigue by modification of the limit value to a safe one.

 

[TGS4]

If this is for a "can under internal pressure", does that mean that it falls under the auspices of ASME or the PED or BS5500? If so, there are rules in these codes and standards.

Weld sizing isn't an FEA task - you should be able to calculate it using hand calculations.

 

[ishvaaag]

Just to point why then my answer, because it seems they did the hand calculations and want more. Normally certainly the thing could be done almost by hand with some simplifying assumptions ordinarily made, but not knowing the case, who knows ... I am very aware of there exists what is to me "infinite" technology out of my reach and present understanding.

 

[EdR]

ISHVAAAG:

Just curious....if the model (and part) are axisymmetric why would you model it as a 3-D part????

Ed.R.

 

[corus]

Modelling the weld would be no good as you'd end up with a stress distribution across the weld, where people are just looking for a single stress. If they're not happy with a hand calc then model it with axisymmetric shells and from the results get the bending moment and direct force at the junction/weld.

PS You don't need a 3D model and Von Mises stress isn't used for fatigue but assessment against yield.

corus

 

[barrind]

In Abaqus it is possible to display a resultant force/bending moment on a cut plane in the viewer module. Could this be used to give the same discreet force and bending moment if you did model the part as an axysymmetric with welds and chose the cut plane from the root through the throat? Then use these results to calculate a single stress based on the weld geometry? I'm not ruling out doing it as a shell model its just that the bottom of this thing is fairly complex, essentially it will have double skinned baseplate with some contact going on, both of which are welded to the side wall of the can. One plate is 10mm thick and the other 40mm. The side walls are 10mm thick on a 457mm diameter.

 

[ishvaaag]

If axis-symmetric I most likely would do initially some kind of hand calculation. Anyway as one goes closer and closer to the details one always feels somewhat uneasy about oneself's limits of knowledge (mine's in FEM are more vast that I would love) and so I understand well the worries of the people that have passed this detail to another colleague more knowledgeable in the field. So when I talk of making a 3D model, well, it is for when I also would love to understand more on what happening. Assuming that one does well, it could extract sectional behaviour ... it is not such more complex behaviour what they must be seeking at the moment?

Respect fatigue, well, yes ... current practice goes along the line of identifying the detail as pertaining to a class, then the intricacies of the check are selfcontained and black-boxed to whatever stated about in the classification process. You are right, in another words.

However essentially what it is being done is a reduction on some limit stress or range of, and there may be cases where the details may not be directly amenable to one classed detail. So for critical applications you would modify the detail to one in the books, for other test and would ask for a class, and for others of lesser concern it might be agreed that a reduction of stresses limiting the overall behaviour might be useful.

 

[corus]

In an axisymmetric shell model just select the sidewalls, for instance, and plot the resultant force, and moments. Normally these would be zero at all nodes within the model as they must balance, but selecting just the elements that exclude the base plate will give you the element nodal forces at the weld. Check the force as it may be in units of force per circumference, ie. N/mm, and moments will be in units of Nmm/mm or N. It'll say in the manual somewhere.

Looking at the stress in the throat of the weld in a full 2D model may be confusing, as I think you'll still have some kind of stress distribution going on. Personally I'd run a simple axisymmetric model that you can definitely check by hand, and compare methods of evaluating the stress in the weld throat by FE just to be sure.

corus