Proof Test vs. FEA 1

[steris]

Hi All,

Short Version:
For a Section VIII Div 1 vessel, does a UG-101(m) burst test generally allow for a higher MAWP than FEA would?

Long Version:
My company has a number of Section VIII Div 1 vessels that were originally designed and certified using a burst test as per UG-101(m). We are currently working on a new design and were using FEA to avoid the cost of creating a vessel for burst testing. The new vessel is fairly similar in design, load, and construction to our older vessels. However, a linear elastic analysis showed that our stresses were way off of allowable. We ran a non-linear plastic analysis and saw that we only needed a little of strain/plastic deformation to bring everything within the allowable stress. For kicks, we decided to analyze our older designs that were certified with burst test using a linear elastic model. Sure enough, our older designs that passed the burst test with flying colors failed by finite element analysis. These results suggest that FEA is much more conservative than a burst test is, however intuitively this doesn't seem right. Mechanically I understand that allowing for some plastic deformation allows the stresses to redistribute - a process that occurs during a burst test but is not permitted in a linear elastic analysis. Am I overlooking something fundamental here or is this the intent of the code?

Any light shed on this topic would be greatly appreciated!

Best,
Steris

 

[TGS4]

We ran a non-linear plastic analysis and saw that we only needed a little of strain/plastic deformation to bring everything within the allowable stress.

Your reference to stresses (particularly allowable stresses) indicate to me that you don't understand non-linear analysis. Have you actually performed an elastic-plastic analysis in accordance with ASME Section VIII, Division 2, Part 5, Article 5.2.4?

Is your elastic analysis actually done in accordance with ASME Section VIII, Division 2, Part 5, Article 5.2.2? It is permissible that the stresses, at the design pressure, will actually exceed the allowable stresses - do you understand this?

In my opinion, the only way to compare burst test with FEA is to use elastic-plastic analysis. However, based on what you have written, I doubt that you actually are performing the E-P analysis correctly. May I recommend that you find someone who is experienced and skilled in applying FEA for pressure vessels to assist you.

 

[steris]

Hi TGS4 -

Thanks for the feedback!

We actually are using an outside firm for all the FEA and final approval but we are doing the initial design in house. Since we are certifying the vessel to Div 1, our limitations are the allowable stresses with the 1.5x factor for local stresses. The plastic anaylsis was performed strictly as a sanity check and will not/cannot be used for design validation - as far as I know there is nothing in Div 1 would allow this.

Your comment that a burst test is most comparable to a plastic analysis seems to suggest that a burst test would allow a higher MAWP than a Div 1 elastic analysis would. Does this sound right?

Best,
Steris

 

[TGS4]

The 1.5 factor that you speak of with respect to local stresses - are these membrane stresses that you are referring to, or are they total stresses?

 

[steris]

The 1.5 factor is for membrane + bending.

 

[zdas04]

Not to throw cold water here, but MODELS NEVER PROVE ANYTHING. If you offered me an untested vessel that passed FEA I'd not only throw you out, I'd burn your business card.

Models are great for pointing out weaknesses that need to be beefed up prior to a physical test. They are really good for finding design bottlenecks that can be corrected prior to putting welder to metal. BUT, at the end of the day you have to build the vessel and break it.

I'm seeing a lot of areas where people are replacing testing with FEA and other computer simulations. Sometimes with disastrous results. When you call modeling "proof" you get bad engineering, bad science, and horrible regulations.

David Simpson, PE
MuleShoe Engineering

http://www.muleshoe-eng.com

"Belief" is the acceptance of an hypotheses in the absence of data.
"Prejudice" is having an opinion not supported by the preponderance of the data.
"Knowledge" is only found through the accumulation and analysis of data.

 

[racookpe1978]

Concur - STRONGLY!

"Models" only shown what "ought to happen" - in a perfectly modeled world with a perfectly modeled vessel with perfectly modeled sub-assemblies (heads, nozzles, joints, rolled shells, lugs, etc) using perfectly assumed (modeled) welds and perfectly uniform steel/joints/reinforcements perfectly and uniformly stressed.

Real world ain't that. Unless you KNOW absolutely and exactly that you exactly and perfectly modeled all the (unknown) flaws and problems in the newly built vessel. 8<)

What is likely to have happened is your previously built vessels accepted the high certification pressures, stretched under the real-world stresses with some strain, but accepted the internal (test) pressure and the new (stretched) geometry without leaking, and went on their merry way receiving (lower pressure) regular stresses without complaint.

You just never knew it.

 

[rhg]

zdas04 is 100% correct

Regards

rhg

 

[TGS4]

As a full-time practitioner of FEA for pressure vessel design, I will also concur with zdas04. However, it is my opinion that the OP's FEA is incorrect to start with.

I have compared several of my FEAs to actual results (be they test results, burst - accidental or intentional, other failures, etc). When you do things right, they will match what really happens. After that, when you validate (important word here) your model, then you can use some engineering judgement to judiciously extrapolate from your validated case to the case of interest.

 

[steris]

Hi All,

Thanks for the input. My initial thoughts were along what racookpe1978 had suggested. I believe our vessels may have experienced some permanent strain during the burst test which allowed for an appropriate MAWP. This seems to be supported by the fact that the strain/deformation measurement requirement UG-101(h) is specifically not required for burst testing. From this, I believe my initial assumption was correct - linear static analysis will not provide adequate approximation for a burst test.

I'll be sure to push back hard on management regarding performing an actual burst test. Thanks for all of the help!

 

[Karloss12]

I'm not familiar with bursting tests, but if you are testing something to failure then doesn't that establish the highest possible MAWP? Expecially as the materials actual strength can be in the region of 30% more then that specified by ASME II.

Wouldn't any other method including a competently conducted elastic-plastic analysis have built in safety factors and use allowable stresses that are lower then the actual physical material?

Or does the B/4 in UG-101(m) typically place the MAWP less then an elastic-plastic FEA result?

 

[steris]

Hi Karloss - In essence that was the core of my original question. It it now apparent to me that an linear elastic model is much more conservative than a burst test. I was actually not aware that an elastic-plastic analysis could be used for Div 1.

 

[Karloss12]

Okay, I have found with my limited use of linear FEA that it's stress catagorisation and lineaisation rules are a bit "one size fits all". It does a great job when conducted competently but as it doesn't model the non-linear stress/strain beyond yield accurately it is a bit clunky and has allot of inherent safety factors built in.

Where as Elastic-Plastic analysis models the problem more realistically.

zdas04, If the old vessels were modelled accurately using elastic-plastic FEA, could this experience be used to design the new similar vessels without the need for a burst test? Even if a case study of the old vessel analysis was included in the calculations?

 

[zdas04]

Models inform investigation, they do not replace it. If the code calls for a burst test (and I don't know the circumstances where it does) then you have to physically break something. If the code calls for a pre-production burst test of a sample of a particular design, then for certain minor modifications of the design FEA may be a reasonable substitute to breaking another vessel (e.g., if the model of the original test matched measured parameters within specified tolerances and the FEA of the modification showed the same stress/strain profile as the unmodified vessel then you could probably make the argument that a burst test of the new design is not necessary).

I've always found the key to success in this kind of thing to be documentation. If I made the call to not do a burst test that seemed to be required, I would write a memo to file explaining my reasoning and attach copies of everything. That approach has stood up to audit for some pretty bizarre deviations.

David Simpson, PE
MuleShoe Engineering

http://www.muleshoe-eng.com

"Belief" is the acceptance of an hypotheses in the absence of data.
"Prejudice" is having an opinion not supported by the preponderance of the data.
"Knowledge" is only found through the accumulation and analysis of data.