Simulation of Nozzles for Fatigue (ASME VIII-2)

[marty007]

Hello,

I'm looking at doing a fatigue analysis for a pressure vessel in accordance with ASME VIII-2. As such, I'm most interested in determining peak stresses, particularly around nozzles, supports, and attachments.

The nozzles for this vessel would be attached as detailed in ASME VIII-1 Fig. UW-16-1(c) or ASME VIII-2 Table 4.2.10 Detail 3 (set-in nozzle, full pen to shell c/w external fillet weld). If I were to recreate the ASME VIII-1 detail in Solidworks Simulation, I would model it as follows:
[ul]
[li]1/8" fillet to model grinding the inside corner[/li]
[li]equal leg chamfer to model the external fillet weld[/li]
[/ul]

This however creates an issue when trying to determine the peak stresses for the fatigue analysis since a chamfer has sharp corners at the toes. These sharp corners would create a singularity in Simulation, and I would obtain higher and higher stress results as I make the mesh smaller and smaller. If I could exempt this vessel from fatigue analysis using the ASME VIII-2 exemption criteria, none of this would be a problem, but that's not an option. How do some of you handle this singularity issue?

Some thoughts that I've had were:
[ul]
[li]Add a small fillet at the toes of the chamfer (how small is appropriate?)[/li]
[li]Model the fillet as an equal leg fillet rather than a chamfer (this would represent less overall weld material, is this conservative?)[/li]
[li]Another method of back-calculating the peak stresses? Does anyone have recommendations?[/li]
[/ul]

I've used a nozzle attachment as an example, but the issue of modelling fillet welds for fatigue analysis is true for all attachment fillet welds (vacuum ring, lifting lugs, etc...).

Thank you for any help/suggestions you can provide!
Marty

 

[badboggs]

Hi Marty,

Welds in simulation is always a problem. Mostly because you have issues with actual welds vs what we model. I remembered that Solidworks used to have a set of Simulation Tutorials by Vince Adams. I'll attempt to attach the weld one here. The whole series is a great read if you can find them online somewhere.

http://files.engineering.com/getfil...aab-489b-ba23-5f960eb727e2&file=116_Welds.pdf

I also had this other discussion of welds. Not sure where I got this one.

http://files.engineering.com/getfil...4a2b-97c2-1e94dec64dd0&file=weld_stresses.ppt

Good luck.

 

[marty007]

Hey badboggs,

Yeah I've seen a lot of discussion on welds in general, but it's been a bit tougher to find stuff in relation to fatigue analysis of welds.

I've been doing a lot of digging around on this subject to try and figure out how best to perform this analysis. There are a bunch of different methods to determine an appropriate peak stress, some of which would be brutal in terms of meshing requirements and run-times.

I dug up a paper by the International Institute of Welding titled "Recommendations for Fatigue Design of Welded Joints and Components". In this paper it discusses a method of surface extrapolation in which stresses on the surface of the base plate are taken at varying distances from the toe of the weld, and the stress is extrapolated back to the toe of the weld using either a linear or quadratic formula.

Does anyone have familiarity with this method and know whether this type of extrapolation is widely accepted as an appropriate method for determining peak stresses per ASME VIII-2? In particular, these calculations may need to be submitted to ABSA (Alberta boiler branch) for review, so I need to know if this is a common approach.

Thank you,
Marty

 

[corus]

In the British Standard BS7608 the method is to consider only the nominal stresses away from the weld and the apply a classification to that stress depending on the type of weld and the principal stress direction to that weld. The method of obtaining the nominal stress in a varying field is, as you say, to extrapolate the stress distribution up to the weld but not to model the weld geometry as that would obviously include the stress concentration at the weld toe, which theoretically would be infinite. Take a look at actual welds to see why that would be the case. You just can't guarantee that an actual weld would be of the perfect shape that you model in your analysis and in addition an actual weld would include defects that you can't model. That's why the fatigue assessment is based upon a statistical probability of failure, and isn't an exact science. The ASME approach is probably the same.

 

[loki3000]

lookup the relevant EUROCODE, FEM or similar US standard for pressure vessels.
pressure vessels and welds have to be checked analytically using the correct code, as they have to adhere to relevant standards (safety, etc).
relying only on FEA data and yield strength is a recipe for a potential disaster and litigation.

http://en.wikipedia.org/wiki/Pressure_vessel#List_of_standards

 

[marty007]

loki3000,

As I stated, we are performing our fatigue analysis in accordance with ASME VIII-2, which is the applicable code in this situation.

The issue is that ASME VIII-2 does not explicitly explain how to determine peak stresses at weld toes, so I am asking how others approach this area of analysis.

Do you have any recommendations?

Marty

 

[greauxpete]

I have seen methods used where the stress is linearized at the toe of the weld (area of max stress),and the membrane + bending stress is used to calculate the number of stress cycles. The mesh was checked for sufficient convergence at the toe of the weld by making sure at least one element on the stress classification line (linearization line) was less than 5% error.

Another method include the hot spot method where points are taken a distance from the toe and the stress is curve fit and extrapolated to the toe of the weld and the stress is then used to determine the cycles. You mentioned this method I believe.