where expertise comes together - since 1996

Web's Largest Portal of Nondestructive Testing (NDT)
Open Access Database (Conference Proceedings, Articles, News), Exhibition, Forum, Network

All Forum Boards
Technical Discussions >
AUT, ECA and Linear Interpolation
Career Discussions
Job Offers
Job Seeks
Classified Ads
About NDT.net
Articles & News

1178 views
06:23 Jul-20-2004
Ed T.
AUT, ECA and Linear Interpolation

I would like to hear from those of you who are involved with AUT and ECA acceptance criteria or anyone with an opinion on it. Has anyone ever thought of or played with the idea of using linear interpolation when evaluating AUT indications using ECA? Does anyone currently use it in the industry?
Let’s face it. The purpose of using ECA is to reduce the number of unnecessary repairs based on actual flaw length versus height. So why not introduce linear interpolation? This would allow even more cost savings for the pipeline industry. I have seen many indications repaired that really didn’t need to be.

For example, let’s assume your acceptance criteria states the allowable length for indications from 2.5mm to <3.0mm in height is 123mm and indications from 3.0mm to <3.5mm is 106mm. In my opinion, if you are allowed to have an indication with a vertical height of 2.9mm with a length of 123mm, you should be able to accept an indication that is 3.0mm in height with a length of 119.6mm and not 106. This is where interpolation comes in.
I would like to get others’ opinions on this.



 
09:32 Jul-20-2004
Isotope75
Re: AUT, ECA and Linear Interpolation How did you come up with the length for that height defect? I am not sure how this "interpolation" is determined.

Explain how linear interpolation works. I have never heard of it. You sound like you want to re-engineer the ECA criteria on the job. If you want to change the ECA you should become an engineer and crunch numbers for a pipeline company.

The AUT operators don't really give a shit about it. If it passes it passes,if it doesn't it doesn't.




 
00:05 Jul-20-2004
Ed T.
Re: AUT, ECA and Linear Interpolation ----------- Start Original Message -----------
: How did you come up with the length for that height defect? I am not sure how this "interpolation" is determined.
: Explain how linear interpolation works. I have never heard of it. You sound like you want to re-engineer the ECA criteria on the job. If you want to change the ECA you should become an engineer and crunch numbers for a pipeline company.
: The AUT operators don't really give a shit about it. If it passes it passes,if it doesn't it doesn't.
------------ End Original Message ------------

My friend, you sound like an AUT Operator. I¡¯m so sorry! I used to be one, but it was making me too aggressive :). I don't expect AUT operators to give a shit, and I am not trying to re-engineer anything. In fact, I am not even trying to convince anyone of anything.
I am only asking for peoples' thoughts on the subject in order to see what people think about it. Stir up a topic.

The way "Linear Interpolation" works is that it takes 2 known values and mathematically calculates the values in between these known values as a gradual curve rather than a step like approach, which is commonly used now. I think it is a more realistic approach to flaw acceptance.
If you look at a typical ECA acceptance criteria, you will see that each range of flaw heights has a specific allowable length.(For example: 123mm§¤ for 2.-<3mm Height, and 106mm§¤ for 3.0-<3.5 height) OK, so if you have a flaw that is 2.9mm, you can accept it if it has a §¤of 123mm. and if you have a flaw that is 3.4mm in height you can accept if it is 106mm§¤ only. Now, if you have a flaw that is 3.15mm in vertical height(the middle of the range for example only), you should be able to accept it if it has a §¤ of 114.5mm, and not reject it at >106mm§¤.
That is all I am saying, and that, my friend, is Linear Interpolation.
There is no number crunching. It is a mathematical fact. But, thanks for your comments. I hope I explained it enough to you. If not, I'd be glad to help you to understand it. It is not that complicated.



 
08:09 Jul-20-2004

Uli Mletzko

R & D, Retired
Germany,
Joined Nov 1998
89
Re: AUT, ECA and Linear Interpolation ----------- Start Original Message -----------
: I would like to hear from those of you who are involved with AUT and ECA acceptance criteria or anyone with an opinion on it. Has anyone ever thought of or played with the idea of using linear interpolation when evaluating AUT indications using ECA? Does anyone currently use it in the industry?
: Let’s face it. The purpose of using ECA is to reduce the number of unnecessary repairs based on actual flaw length versus height. So why not introduce linear interpolation? This would allow even more cost savings for the pipeline industry. I have seen many indications repaired that really didn’t need to be.
: For example, let’s assume your acceptance criteria states the allowable length for indications from 2.5mm to <3.0mm in height is 123mm and indications from 3.0mm to <3.5mm is 106mm. In my opinion, if you are allowed to have an indication with a vertical height of 2.9mm with a length of 123mm, you should be able to accept an indication that is 3.0mm in height with a length of 119.6mm and not 106. This is where interpolation comes in.
: I would like to get others’ opinions on this.
------------ End Original Message ------------

I really have a problem to understand UT specifications or acceptance criteria, which are dealing with flaw size values with an accuracy behind the millimeter decimal point. In my opinion it is very ventured to claim such an accuracy when applying an UT procedure or technique, either manually or AUT, e.g. to circumferential pipeline site welds of ferritic carbon steel, wallthickness about 7 mm to 20 mm, roots not grinded.
I would like to read the performance demonstration reports (possibly blind tests) and to see the metallographic verification and the POD and sizing diagrams, and to learn about the UT background of the persons which are releasing those acceptance criteria.

Regards
Uli Mletzko
MPA, University of Stuttgart, Germany


 
00:17 Jul-20-2004

Ed Ginzel

R & D, -
Materials Research Institute,
Canada,
Joined Nov 1998
1196
Re: AUT, ECA and Linear Interpolation Ed T.
It is interesting that you raise the question of sizing again. Linear interpolation has been a reasonable attempt to accommodate the "requirement" for determining vertical extent of flaws in AUT when using the zonal technique. Like ALL amplitude sizing methods it is a non-precise method. AUT sizing trials have generally been "confidential" and this has led many of us to doubt the vericity of the claims (as Uli correctly raised concerns for!). Very few Round robins have been carried out and some were limited in scope. One carried out by Denys in University of Ghent was limited to 2 companies and these showed that the sizing was not much better than a scatter of guesses (even when one company was using linear interpolation and proprietary sizing algorithms...they had undersizing of up to 5-6mm and oversizing of even greater values).

A good study was published in Insight by Lewis Morgan Nov. 2003. This compared 7 companies' performances on detection and sizing.

I found this study most informative. It was truly a blind test and no company was allowed to investigate the samples outside the test venue. When all was completed the overall results resembled a true Gaussian distribution. This leads one to wonder what all the hype is about with claims for super sizing capabilities. Companies have spent a fortune qualifying their systems based on ability to detect and size flaws that are subsequently sectioned and size estimates compared to size in the macro. What is worse, they are made to spend this "qualification money" over and over again!!!

Morgan's study indicates that, statistically, there is no significant difference between systems provided they use the same sort of sensitivity and zone treatment. With standard deviations in sizing between 1.1 and 1.8mm and Average sizing between -.38 and +.36mm in the Morgan study, we see that when the statistical "null hypothesis" is applied there is no significant variation in results. And variation in gerenal could be explained by random sampling from the same population.

As a guide to how this all stacks up compared to generally accepted results we can look to 2 Codes; BS PD6493 and ASME Section XI. PD6493 was the original guide that Canadian engineers looked to when developing ECA criteria for pipeline girth welds in the 1980s. That document indicated that for automated UT systems a reasonably good Standard Deviation was achieved compared to manual UT. Mechanised systems then were coming in with StdDev of 2 to 5mm. ASME Section XI is the nuclear Code for Inservice Inspection of Light Water Pressure Vessels. When developing a UT technique for qualification to Appendix VIII in that Section the sizing of flaws is required to fall within an RMS of 3.2mm (RMS is approximately the same as STD DEV). When we look at our AUT StdDev on zonal discrimination techniques we are doing much better, with most results falling just over 1mm.

Linear interpolation is a nice simple way of treating amplitude information (which may not neccesarily be the bestoption for sizing, but it is fast and, when within about 1mm it can be considered reasonably accurate). Its development came about as people tried to impose the interaction rules on flaws detected in a weld. e.g. in API 1104 a flaw below the surface must be separated from the nearest surface by a ligament greater than half the flaw height otherwise it is considered to be surface interacting and the height is then calculated as the flaw plus ligament height.

Regards
Ed

----------- Start Original Message -----------
: : I would like to hear from those of you who are involved with AUT and ECA acceptance criteria or anyone with an opinion on it. Has anyone ever thought of or played with the idea of using linear interpolation when evaluating AUT indications using ECA? Does anyone currently use it in the industry?
: : Let’s face it. The purpose of using ECA is to reduce the number of unnecessary repairs based on actual flaw length versus height. So why not introduce linear interpolation? This would allow even more cost savings for the pipeline industry. I have seen many indications repaired that really didn’t need to be.
: : For example, let’s assume your acceptance criteria states the allowable length for indications from 2.5mm to <3.0mm in height is 123mm and indications from 3.0mm to <3.5mm is 106mm. In my opinion, if you are allowed to have an indication with a vertical height of 2.9mm with a length of 123mm, you should be able to accept an indication that is 3.0mm in height with a length of 119.6mm and not 106. This is where interpolation comes in.
: : I would like to get others’ opinions on this.
: I really have a problem to understand UT specifications or acceptance criteria, which are dealing with flaw size values with an accuracy behind the millimeter decimal point. In my opinion it is very ventured to claim such an accuracy when applying an UT procedure or technique, either manually or AUT, e.g. to circumferential pipeline site welds of ferritic carbon steel, wallthickness about 7 mm to 20 mm, roots not grinded.
: I would like to read the performance demonstration reports (possibly blind tests) and to see the metallographic verification and the POD and sizing diagrams, and to learn about the UT background of the persons which are releasing those acceptance criteria.
: Regards
: Uli Mletzko
: MPA, University of Stuttgart, Germany
------------ End Original Message ------------




 


© NDT.net - The Web's Largest Portal of Nondestructive Testing (NDT) ISSN 1435-4934

Open Access Database, |Conference Proceedings| |Articles| |News| |Exhibition| |Forum| |Professional Network|