Fatigue analysis, PD5500 vs Div 2 1

[jtseng123]

Hi dear experts, it must have been discussed before, but please excuse me one time.

Can you tell me the differences for fatigue analysis between PD 5500, which I am not familar with, and Div. 2 ?
Which is more accurate to predict fatigue life ? Which is more conservative ? What are the pros and cons between the two ?

In 2008, I have a vendor provided analysis by both codes.Div 2 came up 200 cycles, and PD 5500 came up 1900 cycles. The person doing it is a PhD, and I can only trust what he did is correct. But the big difference in cycles was shocking me.

So my last question is: are both codes getting close now by the latest versions ?

 

[TGS4]

Which fatigue analysis method in Div 2 was this comparison done for? Are you discussing smooth-bar fatigue or weldments?

 

[jtseng123]

TGS4,
I am confused the term "smooth-bar" used, although I saw it everywhere. Why saying "bar" ? I just saw so many smooth curves here and there.
New Div 2 has weld reduction factor added in. If that was what's lacking in the past and now makes it perfect, then I am confused the question you are asking me.
Anywhere I shall study first to make myself clear ?
Material and weld can not tell whether you are using Div 2 or PD 5500 to predict its life under the same operating condition. So my original question is boiling down to the comparisons of apple to apple, orange to orange between the two codes.

 

[TGS4]

OK, I think that I understand your confusion. In the ASME Code, there are two methods for the fatigue of weldments. The first uses the same fatigue curves that have been in Div 2 since its inception - these curves are based on fatigue tests on smooth bars (round bars, actually) in rotating three-point bending tests: hence the name smooth bar curves. To adjust for the reduced fatigue performance of welds, weld fatigue strength reduction factors (FSRF) were generated. In Div 2, these FSRFs are based on the weld geometry AND the level of inspection. In PD5500, these FSRFs are based on weld geometry ONLY.

New in the 2007 edition of Div 2 was the Structural Stress Method for the Fatigue of Weldments. This method is based on tests of WELDED joints. Hence, it is fundamentally different from the smooth bar tests described above.

I guess that you are really comparing the FSRFs between Div 2 and PD5500. Both are somewhere between crude approximations and wild a$$ guesses. That you are within an order of magnitude between these two methods is actually expected - especially in the low-cycle regime.

I would strongly (and I mean STRONGLY) suggest that if it is anything important, use the Structural Stress method. I have been using it for 8 years (yes, before it was officially published) and I can say that for failure analyses, it is frighteningly good. In the low-cycle regime, I only use it.

 

[Karloss12]

The S-I curves in PD 5500 are for 2 standard deviations (for instances where additional inspection is not required).

ASME VIII Div 2 uses 3 Standard Deviations as default. So 200 cycles / 0.6 ~= 333 cycles for PD 5500. (0.6 roughly converts 3 SD to 2 standard devs.)

Is the 200 cycle / 1900 cyle example above for a fillet weld of a Nozzle Reinfocing Pad? For ASME if you can't inspect the root of the weld than RSRF has to be 4. PD 5500 doesn't request this. So 333 x 4 = 1333 cycles is probaly equivilant to PD 5500. It's in the ball park. (just for the fillet weld of a nozzle reinforcing pad.)

 

[TGS4]

Karloss12 - I agree with your assessment of the SD approach in PD5500. However, the standard ASME S-N curve does not acknowledge any sort of statistical treatment. It is not 3SD below a mean, it has a factor of 2/20 on the mean curve.

Furthermore, while I agree with your assertion about the FSRF (Fatigue Strength Reduction Factor) needing to be 4 in VIII-2 vs PD5500, your math is incorrect. That FSRF is applied to the STRESS, not the CYCLES.

 

[Karloss12]

Thanks TGS4,

Dividing the CYCLES by FSRF is wildly wrong.

Where do you get the 2/20 from for the smooth bar method?

I assumed that because the Welded Fatigue curves of Div 2 use standard deviations (3 SD as deault), then the smooth bar curve is good for 3 SD.

 

[TGS4]

Karloss12 - the 2/20 comes from as far back as The Criteria Document, but can also be found in ASME PTB-1 in 5.5.3.2.

Unfortunately, ASME decided to use a best-fit curve, adjusted for mean stress, and then apply a factor of 2 on stress and 20 on cycles, and take the cumulative lower curve. (This explains the kink in the curve itself at ~1e6 cycles - cycles takes over from stress at that location)