Nozzle fatigue analysis in Div 2 1

[jtseng123]

Dear all,
I have mechanically cyclic loads (axial, bending, torsion and shear from a rotating agitator) acting simultaneously at a nozzle. The design cycles are specified. (10ex9)
(1) shall the fatigue analysis be done by calculating stress (by FEA) from each load case, find out its cycles from fatigue curve, then use Accumulated Fatigue Damage per Eq 5.46 to sum up all 4 load cases ? or
(2) just run stress analysis with all loads acting simultaneously, then find out the cycles and compared with the design cycles ?

The vendor uses (1), but I don't think that is correct because it ignores the stress tensor. It assumes stress generated by each load case is acting in the same direction, although that is the worst scenario to sum everything up positively. However, when all 4 loads acting simultaneously, the maximum stress intensity may be reduced due to stress tensor that stress subtraction occurs.

 

[LSThill]

jtseng123 (Mechanical)

see attach

ASME Section Vlll, 2 rules published 2007 and later for load cases and stress calculation per ASME PTB-3-2010 Example Problem Manual PART 5 Page 5-2,-,5-59

 

[jtseng123]

LSThill, thanks. but I can not figure out the answer from the link. I am not an FEA guy, but I do need to review vendors' FEA once a while and when vendor giving something odd against what I have experienced, I can not say it is wrong unless I have a good technical basis.
Let me rephrase my original question, making it easy to understand:

Each nozzle has 6 external forces (Fx, Fy, Fz, Mx,My, Mz)acting at the same time. When we do local stress analysis, either WRC107 or FEA, we put all loads together and find out the maximum stress at a point. We never separate each force to find out its own stress and then add or subtract all 6 stress together.

When doing fatigue analysis, shall all 6 forces be applied together and find out only one maximum stress and get only one number of cycles from the S-N curve ? or get 6 maximum stresses from the 6 forces, and find out 6 numbers of cycles from S-N curve, then apply Linear Damage Rule or Miner's rule to sum up all 6 damage factors ?

 

[jamesl]

I think the vendor was misusing the Miner's rule. If the load cases are indeed acting simultaneously, what the vendor was doing would underestimate the damage. A 200 ksi alternating stress intensity would do much more damages in 100 cycles than (2) 100 ksi stress intensity acting consecutively in 200 cycles. Eq 5.46 shall be used when they are different types of stress cycles. If the agitator creates axial load for, say a million cycles, then shear load for two million cycles, etc., it would be right to use Eq. 5.46 to assess the damage.

 

[jtseng123]

jamesl, all 6 loads (3 forces, and 3 moments) generated by agitator are happened simultaneously, just like a clock keeps changing and back again in one cycle. All loads will have the same design cycles in the life of the agitator.

I thought the use of Miner's rule or Eq 5.46 shall be the same type of load , but different intensity and frequency.

I know many people have done fatigue analysis for similar case due to external nozzle loading, and I am just want to know whether the alternating stress and fatigue cycles shall be calculated by applying all loads together, which making sense to me, or do one by one as my vendor is doing, which I think is too conservative due to linear addition of Miner's rule.

 

[TGS4]

jtseng123 - glad that you caught the problem and asked a question here. If the loads indeed are simultaneous, then as jamesl says, they should be analyzed concurrently (in the same analysis). My biggest concern with these types of loads is that you will have stress reversals (which can occur with load reversals). It is very important that the stress ranges be calculated at the component level when calculating the stress ranges.

In my opinion, your second option would be more appropriate. However, make sure that the entire cycle is evaluated, including load reversals. If you miss that, and you only capture half the stress, your life will be affected by a factor of 20.

BTW, in this high-cycle regime, how are the analysts treating the welds?

 

[jtseng123]

TGS4, thanks.

The agitator supplier did provide maximum and minimum load cases in a cycles with proper force directions. Some forces, such as shear and bending moment, do show reversals. By applying FEA with all 6 forces together, it will be very easy to get maximum and minumum von Mises stress for each cycle at a specific location.

I got much clear picture now. Miner's rule shall not be used in this case. Vendor shall find out 6 stress components from both load cases at a specific location, then follow the code. But it is extremely more complicate in the new code than the old code to find out the alternating stress. I can only wish vendor's FEA program, SolidWorks Simulation, will do it right.

The nozzle is integral reinforcement, full penetration weld without repad. It there a concern in the analysis ? Weld was modeled without any difference to nozzle or head. Code does have a section talking about fatigue in weld. It is so complicate that I do not know how to relate to the analysis or ask vendor what they shall do. Please help.

 

[TGS4]

First things first - stop thinking about fatigue (at least in the context of the ASME Code) with regards to stress invariants (von Mises or Tresca). Calculate the stress ranges at the COMPONENT level (σ11, σ22, σ33, σ12, σ13, and σ23), THEN roll those stress components up into a stress range invariant. Ask your analyst to show you how what they are doing complies with this. Suffice it to say that if they are calculating von Mises stresses at each point in the cycle, then likely they are doing it wrong.

BTW, it is NOT more complicated in the new Code vs. the old Code - the fatigue rules are identical with the exception of using von Mises as the invariant vs. Tresca. The approach of calculating the stress rages at the component level and then rolling that up into a range invariant is identical.

As far as the nozzle weld goes, yes, by not doing anything, they did something wrong. Because you are in the high-cycle fatigue regime, I would not recommend the structural stress method (there is insufficient data in the high-cycle regime, so you would be extrapolating from the existing data, which we all know is dangerous...). However, the Code provides you with another method, the FSRF (Fatigue Strength Reduction Factor) method - 5.5.3.2, Step 4, a), and Table 5.11 and 5.12. There, the stress penalty factor depends on the level of weld surface improvement (machined vs. as-welded) and inspection. For the weld that you describe, you could have a FSRF of 1.2 to 4.0...

 

[jtseng123]

TGS4, thanks again and I will follow up with vendor regarding the weld analysis.

 

[jtseng123]

TGS4,
After reviewing Code Step 4.a, what does it mean "local notch or effect of the weld is accounted for " or "not accounted for" ? The weld was modeled in the FEA, does it mean " accounted for" ? How a finite element program, such as ANSYS or many others, to take this into account ?
Does it mean vendor shall model the weld separately, and trigger somewhere in the program for the specific Kf factor from Table 5.11 and 5.12 ? Please help again. Thanks.

 

[TGS4]

At the transition from the base metal to the weld cap/fillet, in the as-welded condition, there will be a notch of some description. If the analysis modelled only the ideal geometry, then I would consider that the local notch/weld effect is not accounted for.

In this case (and in all cases when I use this fatigue of weldments technique), I do nothing special to model the weld; all of the additional effort is in the post-processing. Essentially, you need to factor the calculated stress range by the appropriate factor in Table 5.11/5.12.

Glad to help.

 

[jtseng123]

TGS4,
When you do your own analysis, does your commercial program have a place to input Kf for specific elements such as weld areas, and it will include in the post-processing ? or it must be done by hand after FE analysis ?

Like I say I am not a FEA user, but have to review FEA reports. So I need to know some basic about commercial program to make sure this is a typical built-in function.

Many FEA reports I have seen, normally only a description of what will be done and design conditions, then attached bunch of colorful outputs, and conclusion. If the report comes from an experienced consultant that I know of, I can only trust what they are doing is right. If coming from an unknown party, such as my case now, and showing all kinds of good results, I will need to point out something from code to make sure they don't miss it.

 

[TGS4]

In my programs (ABAQUS and ANSYS), there is not an option to input Kf for a weld. It is something that I do manually in the post-processing. The Code was written with the intent that such a calculation would be done manually in the post-processing.

I completely hear what you are saying about FEA reports. I started teaching a VIII-2 Part 5 course, and I dedicate almost 2 hours to the topic of writing the report. I sometimes have the opportunity to write a specification for a client (an FEA Spec, if you will) that is provided to vendors along with all of the other client specs. Both in my course and in my spec, I consider that the most important aspect of the report is that each and every figure (plot) should be referenced in the body of the report itself. If the analyst cannot dedicate two or three sentences to explain the significance of the figure, and to discuss what it shows, then the figure has not reason to be in the report. I especially hate the reports that have almost a page of text, followed by 100 pages of colourful output, finished all with a one-sentence conclusion "The design meets Code."

 

[jtseng123]

TGS4,
The tough part is, Kf only applies to the first term of Eq. 5.30. If applied to the entire equation, that will be simple.
Is it easy to manually do it in your experience? Or is there any spreadsheet readily available to do it ?

(It just comes to my mind that, if this vendor has trouble to do it right, I will ask them to apply Kf (= 1.2 in this case) to the entire Eq 5.30. If it passes, then OK. Or ask grinding weld smooth and do the required NDE so Kf =1.)
Thanks for your help again.