Skirt junction design for Div 2 vessel in cyclic service 4

[jtseng123]

Dear all,

Please refer to the attached sketch. This is for Div 2 vessel in cyclic service. Our joint partner insists on using weld overlay from Div 2 sketch while I see no problem with shop design which has less weld to avoid rubber band or weld harden effect so to keep the junction ductile to minimize stress riser. FEA for fatique analysis will be performed.

(a) Can someone shade the light which is better design to prolong the fatigue life.
(b) I still wonder why it is required to strip the base metal on head and then weld overlay with C.S. weld rod to the C.S. head and with huge weld build up to make up the skirt portion per Div. 2 sketch. It is a lot of work and cost. Has anyone really done that ?

 

[TGS4]

The critical parameter for the fatigue analysis will likely be the radius on the inside of the skirt. Ideally, if that could be guaranteed by using a forged ring instead of the weld, that would be better. Welds are somewhere between worse and much worse in fatigue service. That said, I would tend to leave the exact process up to the fabricator for them to say which approach is easier to guarantee the inside crotch radius.

As far as the fatigue analysis goes, I would likely want to see the results of both an FSRF (Kf) analysis with an FSRF equal to no less than 2.0 (I am not certain of the effectiveness of a MT/PT on that inside crotch radius) and a Structural Stress Method analysis calculation.

I don't understand what you mean by "rubber band" or "weld harden effect". Would you please explain. What is cycling - pressure or temperature or both?

 

[jtseng123]

TGS4,

Rubber band effect is what we describe the girth seam that is not so ductile/flexible such that base metal will bulge more than girth seam, especially in a cyclic coke drum.

The skirt junction will be 100% RT and 100% MT.

Our joint partner did understand the fatigue will pass. It is not the issue the crotch radius or what factors shall be used in the analysis. They just have no confidence with vendor's design regardless how good the analysis is, and want to strictly follow Div 2 Fig 4.2.4 (f) which I don't know if it is better than vendor's design.

Also, I do not know any technical reason why stripping back base metal and put so much weld build up to make up the skirt per Div. 2 sketch. What I know is, the more the weld, the more chance to fail.

From "design" point of view, not from "analysis" point of view since we can all make "analysis" be equal, is DIv 2 sketch better than vendor's design ? and why ?

 

[TGS4]

I'm not sure why weld would be considered less ductile or flexible. It has the same Young's Modulus as the base metal and is required to have the same ductility. There are indeed issues of welds in coke drums, but those relate to differences in yield strength mis-match - because the stresses during operation of a coke drum exceed yield.

If your joint partner doesn't trust the analysis, then I don't know what to say. If the vendor is willing to stand behind their design for the fatigue life of the vessel, then who are you to object?

Why do you think that "design" is different than "analysis"? Aren't you performing a Design BY Analysis?

What I know is, the more the weld, the more chance to fail.

And how do you know this?

 

[jtseng123]

TGS4,

In all, it is the comparison between metal produced by mill which is homogenous and weld which has many external factors involved and it can have higher hardness than base metal.

For sure both designs will be "Design by Analysis". It is just the personal preference that our partner opts for Div. 2 sketch rather than vendor's design. Maybe weld overlay by Div. 2 will have stronger bond than directly welded to head w/o weld overlay. I believe FEA won't tell the difference.

 

[TGS4]

If you are really concerned, a forged ring would be better. Since you're going down the road of a weld, both designs have similar issues with significant weld build-up.

How many cycles are we talking about? Are the stresses above or below yield?

 

[jtseng123]

TGS4,
10,000 cycles, from 80 degF to 540 degF, but not much pressure fluctuation. Also, we have added a hot box that will reduce temperature gradient. It is too late to change to forged ring, and our partner does not trust any result from FEA fatigue analysis and insists on using Div 2 sketch with weld overlay as the remedy. I believe FEA from both designs will show no difference but is Div 2 sketch a better design than vendor's design without weld overlay ? Even coke drum, we don't do weld overlay for the skirt support.

Stress above or below yield is not a concern. It is our partner saying "their many experience and lessons learned " that the vendor's design is no good, regardless the stress analysis.

 

[TGS4]

A hot box won't really help at those temperatures.

I had a mentor early in my career, who had a saying that would be appropriate for this situation - "Show me the math!". It's fine (ok, no it really isn't...) that someone doesn't "trust" an FEA. But to rely only on some vague "experience" is a total cop-out. My reaction is always to reply "Show me the math!". Any reply that isn't based in mathematics and physics, but pseudo-math, pseudo-physics, and a whole lot of hand-waving is simply BS.

You want to know which one is better? Follow the math and physics. Set a criteria and evaluate in accordance with those criteria. Anything less is BS.

 

[jte]

Well, I rarely disagree with TGS4 but I'll put a different perspective on this. Yes, the "show me the numbers" (thanks, Dennis!) attitude is very valuable at times. But there is also some value to tribal knowledge. Not to say that tribal knowledge is always right - I fight the "That's how we've always done it" attitude weekly. But there are things which no reasonable amount of number crunching can forecast.

If one want's to claim that a great FEA will absolutely predict a failure - then please let me know what year my coke drums will experience their first cracks!

Weld material is not the same as plate material which is not the same as forgings. But the vast majority of number crunchers do not attempt to include anisotropic material properties in their models. Some weldments are easier to weld than others. Loadings are rarely truly symmetrical - are we filling this vessel at sunrise while one side is hotter than the other? This is where tribal knowledge can play a role.

I do not allow slip on or lap joint flanges on my vessels. The math, and indeed B16.5, says that they are fine. But tribal knowledge says they are not as reliable as weld neck flanges. In this case, the strict "show me the numbers" approach is BS.

Having said that, in the particular case being discussed, I agree with the sentiment that avoiding massive weld buildup* is better - but that judgment is based on tribal knowledge!

* For those Section VIII committee members reading this, note that the word "overlay" in Div. 2 Fig. 4.2.4 is incorrect. An overlay by definition is a weld deposit of material which does not share the same basic chemistry as the base metal. A weld buildup, on the other hand, is a weld deposit which shares the same basic chemistry as the base metal. See, for example, PCC-2 Article 2.11 part 1.5. Or Div. 2 paragraphs 4.2.5.2, 6.2.4.9, and several others.

 

[r6155]

@jtseng123: you said "It is too late to change to forged ring...." . Sorry but it is sound no seriuos to me. On what were you thinking during design procedure?

I agree with jte

Regards
r6155

 

[TGS4]

jte - some excellent points there, and certainly experience ought not to be put down. However, there is always a physics- and mathematics-based reason behind why everything happens/happened, which is the basis for tribal knowledge. The challenge for us engineers is to discover that basis.

Your example of cracks in coke drums is an excellent one. In that regard though, for such a generally stochastic process (fatigue crack initiation), caused by loadings that are not know a priori, we still do a generally good job of predicting intervals in which cracks are likely to occur.

Similarly, your example about lap-joint or slip-on flanges is another good one. It highlights that the Code (in particular) and our calculation methods (in general) are agnostic with respect to service conditions. Slip-on flanges are less reliable than weld neck flanges because of the presence of fillet welds, which greatly reduce their effectiveness in cyclic service; they have a built-in crevice for hydrogen accumulation; etc. These failure modes are generally not considered in typical "calculations", but when included demonstrate the rationale behind the tribal knowledge from experience.

Based on my experience, the typical reason why experience-based tribal knowledge doesn't sync with calculations, is that the calculations don't consider the full gamut of failure modes - and therefore it is the calculations that are deficient. Which is why I am a strong advocate of using a failure modes approach to calculations. My experience has been that when this approach is followed, and the appropriate potential failure modes are identified and addressed (this step needs to be be the bridge to the experience base), then is always a correlation between the math and experience.

In the absence of a correlation between the math and experience, we engineers end up being reduced to nothing more than tribal shaman, for whom the way things have been and the way things are are nothing more than magic.

(Regarding the terminology error in Figure 4.2.4(f), the wheels are in motion to have that corrected)

 

[r6155]

"The skirt junction will be 100% RT and 100% MT".
Are you sure?. Or will be 100% UT and 100% MT?

r6155

 

[jte]

TGS4-

I agree with much of your post. However, you failed to address one crucial phrase in my post above:

But there are things which no reasonable amount of number crunching can forecast.

Clearly "reasonable" is a matter of perspective! My perspective as an owner/user is a tad different from yours!

More seriously, though, if we as engineers had the funding, time, and resources (i.e. warm bodies) to evaluate everything to the extent which you imply we should strive for, then it is time to substantially drop the design margins in the codes. No need for a factor of 3.5 when you know everything there is to know about materials, fabrication, QA/QC, shipping, installation, and future operations.

 

[TGS4]

Indeed jte. Design margins on ultimate:
[tt]"old" Division 1 - 4.0
"new" Division 1 - 3.5
"old" Division 2 - 3.0
"new" Division 2 - 2.4
..... Division 3 - 1.8[/tt]

A definite progression of doing more results in lower necessary design margins. Nevertheless, you are correct in that I did miss that phrase. And that is indeed where experience plays an important role.

 

[jtseng123]

As of the latest, due to the higher cost per Div 2 sketch to do weld overlay on head and then huge weld buildup on top portion of skirt, our joint partner accepts vendor's design with FEA fatigue analysis. This vessel is PWHT'ed due to the thickness. The junction for sure will be 100% UT if RT is not practical, and 100% MT.