Weld Fatigue Analysis Using BS7608/5400 And FEA 2

[jas165]

This question concerns the computation of stresses for use with the set of fatigue curves given in BS7608/5400.

Given a complicated steel weldment that does not lend itself to any of the classifications set forth in BS7608/5400, the question is how to compute stress near a weld in question that is applicable to the BS SN curves. Since the slope of the weld SN curve is roughly 3, the computed stress is taken to the third power which yields drastic changes in calculated life for even small changes in stress range.

Should the location of stress calculation from an FEA be taken at the weld toe (if modeling welds), where the weld toe would be (if not modeling welds), or something like 2 weld leg distances from the toe?

What type of stress computation should be used? Absolute maximum principal stress, signed Von Mises, or simply the transverse or longitudinal stress relative to the weld line? Or is this not cut and dried; rather a set of guidelines based on the biaxiality ratio will determine the type of stress calculation to use? Would it be prudent to use absolute maximum principal stress for biaxialities between 0 and -1 and Signed Von Mises for biaxialities between 0 and 1?

More on this and other weld fatigue topics later.

 

[truckcab]

I am not that familiar to BS7608 but generally in codes I believe that the stress concentration factor is already icorporated in the classification of the welds. Depending on how accurate you model the weld geometry in your model you will to some extent compute stresses influenced by the SCF. Consequently, since most codes suppose that you use nominal stresses meaning that you are very conservative if you use e.g. nodal stresses at the weld toe.

I'd check the gradient and take the stress a bit away from the weld and transform it into "weld stresses" i.e. perpendicular to and along the weld. Shear stresses projected onto the surface in the smallest section (dont know if u use the "a measure" in your vocabulary).






Live Long and Prosper !

 

[corus]

If you have a weld that isn't easily lend itself to the classifications then I'd err on the conservative side and choose a worst case classification for that weld. You can also find further examples of weld classifications in the Pressure Vessel code BS5500 or its EN equivalent. Also, I think BS7608 has been superceded by an EN code which has a slightly different method of assessment and further examples of classification, I believe.

Generally the stress you use for assessing weld fatigue life is to calculate the principal stress range. In training courses given by members of the TWI, they suggest that from the FE software obtain the difference in the stress components and calculating the principal stresses from that stress range. Use the absolute maximum value of the 3 principal stresses for your assessment. Generally the stress components are obtained by plotting the values up to the weld toe and fitting a curve to that data, excluding the values produced by the geometric stress concentration in your model. If you're looking at a weld at a corner then presumably you'd have to fit a surface to the data, which is a little ridiculous in my book. In that case you're probably better taking some nominal stress range value 2 weld leg lengths away. The direction of the maximum principal stress range will of course play a part in determining the weld classification. You don't use Von Mises stress for fatigue evaluation nor do you normally consider the sign of the stress due to residual stresses from the weld.

There is another method of assessing welds which uses the nodal forces from the model to determine the fatigue life regardless of the type of weld. This was discussed here sometime ago but I've not seen it applied in practice yet.

You could also look at the TWI web site

http://www.twisoftware.com/home/home.jsp

which has some software to use with 10 day trials, and a forum there.





corus

 

[trainguy]

I have heard that the Hot Spot Stress method addresses your concern, but I have not tried it yet.

tg