Estimate of first appearance of cracks

[stryker1080]

Hello,

I was wondering if people can provide some guidance as to various criteria for when one can expect first indications of cracking. Couple of specific questions i have

1) When doing an ASME Div 2 - Part 5 analysis, i understand that this method lets you determine the permissible number of cycles with an accepted factor of safety. Out of curiosity, what is the basis of this method? For B31.3 piping tests by Markl, they did the tests until water leaked (through wall crack) and they applied an arbitrary FOS on stress and cycles (~2 on stress and ~20 on cycle).

Is the same basis done for Div 2 - Part 5? In that case, is it possible to start seeing cracks even if you are only at 5, 10, or 20% of your permissible cycles? Any literature that people can recommend on this subject?

2) Can someone recommend some literature or codes which would reference an approximate estimate for when you would expect to see cracks. I'm looking something along the lines of:

- B31.8 has some guidance as to allowable strain (4% and 6%) used for dent evaluation. Some literature i read suggests that first cracking can be seen around ~ 12% plastic strain. Would performing a linear elastic-perfectly plastic or a plastic analysis and looking at the plastic strain be a good indicator?
- Some reference i read have a simple formula which includes a ratio of principle stresses to von mises (gives you a ratio of the triaxality) and multiples that by the uni-axial failure strain to give you a "new" failure strain.

Just curious what people are using as a criteria.

I am asking more out of curiosity. We came across a situation where we had a local bake out on a vessel and it did not have uniform gradient across a discontinuity. We inspected the vessel post bake out and found no cracks. Contractor says they do this all the time... For interest sake, i did an FEA and ran a thermal analysis followed by stress analysis and found fairly high stresses at the discontinuitiy. Wanted to get a "data point" or gauge what stress level or strain would be a good level for when to definatley inspect or what areas to focus on...

 

[TGS4]

The data upon which the fatigue curves in VIII-2 are based define "failure" as anything from a leaking crack (Markl) to the appearance of 1mm or even 0.5mm cracks. There is not one consistent definition of failure in the data, which makes the handling fatigue difficult.

The design margin between the data and the published curve is not arbitrary - it is the curve resulting from the more conservative of 2 on MEAN (meaning the average of the fatigue data) stress and/or 20 on MEAN (same definition) life. This design margin was deemed sufficient to be a lower-bound to the data.

You are correct about the triaxiality ratio being one definition for when cracking will begin - see Protection Against Local Failure in VIII-2 Part 5.

What is relevant is not the initiation of cracking, but what happens with those cracks and when those cracks become "bad cracks". For that, you need to apply a fracture mechanics approach -

http://becht.com/blog/use-of-fracture-mechanics-for-fatigue-design.

In general, this is the appropriate approach for calculating the inspection interval.

One of the biggest reasons for the large scatter in fatigue testing data is that the initiation of cracks does not appear to be particularly deterministic. No one really knows why.

For fatigue analyses, the regions that typically see fatigue damage (cracks) are welds. And the smooth-bar approach with the FSRF-modifications is probably not the best approach to evaluating fatigue. I would recommend the Structural Stress Method for Fatigue of Weldments - as detailed in VIII-2 Part 5. The technical background to that is well covered in ASME PTB-1 and WRC-523.