Fatigue Analysis

[Sail_Fish]

Hi all,

I am looking at performing some fatigue anallysis (theoretical only) and want to udnerstand the differences between the PD5500 and ASME VIII D2 method a bit more.

With the PD5500 method, you are required to estimate a required number of cycles that then determines which of the S-N curves you use as per Table C.1. If I assume that the required number of cycles is less 10^7, my answers fit 'very' well with the asnwers calculated using ASME VIII for the same vessel. However, if I dtermine that the required number of cycles is greater than 10^7 then my answers calculated using PD5500 increase substantially, whilst my answers using ASME VIII stay the same. To my knowledge ASME VIII does not have require number of cycles as a variable. I understand the reasoning behind the increase in allowable number of cycles due to the change in gradient of the S-N curve, but I am wondering how you factor this into the ASME VIII method? Does ASME not make allowances for 'High cycle fatigue'?

Thank you for any advice.

 

[TGS4]

This is so theoretical, I can't even follow what you're asking. Do you have a specific example that demonstrates your issue?

 

[Sail_Fish]

TGS4 - I am trying to determine the allowable number of cycles for a vessel operating under cyclic pressure fluctuations using the PD5500 method and ASME VIII method fatigue methods..

Within the PD5500 method, the constants 'm' & 'A' are given depending upon the weld class. There are two sets of values for 'm' and 'A'. One set of values if the estimated required number of cycles of the equipment being assessed is lower than 10^7 and a higher set of values if the required number of cycles is greater than 10^7. The vessel I am analysing has a required number of cycles in exceedence of 10^7 so I am required to use the higher set of values for 'm' and 'A'.

I am also running the calculations usign the ASME VIII structural stress method. This method doesnt accomodate for the estimated required number of cycles the equipment will be subject to.

If I assume the required number of cycles is less than 10^7 and as instructed, use the lower set of values for constant 'm' and 'A' I get for e.g 10^9 allowable cycles using the PD5500 method. If I run the same analysis using the ASME VIII structural stress method I get a very similar result.

If however I decide the vessel now needs to have a requried number of cycles of greater than 10^7 I am forced to use the higher set of values for constants 'm' and 'A' within the PD5500 method and the allowable number of cycles now increases substantially for e.g. to 10^11. Because the ASME method doesnt take into consideration the estimated required number of cycles, the answer stays the same.

So for an estimated required number of cycles that is less than 10^7, the PD5500 and ASME method produce very similar results. For estimated required number of cycles greater than 10^7, the emthods give very different results due to the constants 'm' and 'A' within PD5500 changing.

My question therefore is, why does PD5500 take into account the required number of cycles and give a greater fatigue life for 'Higher cycle fatigue' (i.e. for a required number of cycles greater than 10^7) and ASME does not take into account the estiamted required number of cycles? Which method is to be used for higher cycle fatigue?

Hope I have explained it well enough.

 

[TGS4]

In the Structural Stress method, the welded joint data sort of hints at a "knee" where the slope changes, but there just isn't enough data to support changing the slope of the curve. The Structural Stress Method (of which ASME uses the welded joint fatigue data, as do many other industries) simply needs more high-cycle fatigue tests for as-welded joints.

If you really are concerned about high cycle fatigue, the best thing that you can do is ensure that the surface is flaw-free (using whatever inspection technique you deem appropriate), and then use a fracture-mechanics approach. In many environments, there is a threshold stress range that an infinitesimal crack (sized per the threshold of detectability of your inspection technique) simply will not propagate.

 

[Sail_Fish]

Many thanks for the explanation. I will look into the fracture mechanics approach.

 

[DriveMeNuts]

Sail Fish, the reason that ASME hasn't added a change in slope is that the ASME VIII fatigue method isn't as advanced as the PD 5500 Appendix C method, for instance PD 5500 is fundamentally based on Paris law and therefore provides a probability of survival, where as ASME VIII just says yes or no for the smooth bar method.

I'm surprised that you are even able to get similar results between the two methods. I get very different results, especially after applying the ASME Fatigue reduction factors. They have completely different philosophies to the safety margins.

The S-N curves and underlying fracture mechanics studies (BS 7910) show that the Stress-life relationship can be modelled as a linear logarithmic line up until some number of cycles (it has changed from 10^7 to 5×10^7 cycles in the latest PD 5500:2018). Beyond that number, the rate of fatigue damage reduces and therefore the line continues at a different slope. Testing has shown that further off to the right of the graph, the slope is zero for some materials (i.e fatigue doesn't occur).
This is accepted by ASME VIII div 2 specifying that where the stress is less than 20% of the components rated pressure, then fatigue assessment is exempt (i.e. fatigue doesn't occur).

Both ASME and PD 5500 Quality assurance fatigue methods are based on very conservative assumptions, so I don't see any need to move to the very time consuming fracture mechanics method. If your component is not welded and you are not interested in intensive inspecting for cracks, it is much quicker to just use the conservative class C S-N curve and be done with it. BS 7608 (the basis of PD 5500 Appendix C) may even allow you to use the Class B curve if you conduct additional QA.

If you are using ASME VIII then TGS4's suggested approach of using fracture mechanics is appropriate. However with PD 5500 (or BS 7608) it can be avoided as the method is fundamentally based on fracture mechanics.

Flaw free material is the stuff of laboratories (Class A for 7608). Just ensure the minimum flaw size corresponds with the desired probability of success and appropriate Class B or C, then use the S-N curve and you've got your answer. Allot quicker and easier than fracture mechanics.