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
Career Discussions >
Re: NDT Reliability methodology -- other people interested
Job Offers
Job Seeks
Classified Ads
About NDT.net
Articles & News

623 views
07:56 Feb-10-2004

wolfgang Bisle

R & D, Project leader & Group Leader InService Inspect.
Airbus Deutschland GmbH,
Germany,
Joined Jul 2000
35
Re: NDT Reliability methodology -- other people interested

ate you need a high number of samples without discontinuities... a very expensive situation, if you declare only a complete testspecimen to be "one" sample...

Hope I have put the situation in correct words as I am not a "native english speaking expert in statistics", so some expressions used in german for this topic can't be easily retrived from regular dictionaries (I mostly use http://dict.leo.org - a very good living project of the univ. of Munich), sometimes it is hard for us to hit the definition 100%.

Conclusion: I like this discussion you did, very much and maybe it gives us some additional ideas to improve our work to the benefit of safety


--
This message was posted into the NDT.net Forum.
http://www.ndt.net/wshop/forum/forum-2.htm
Please do not reply to email: forum@ndt.net
For your reply to this message go to:



 
01:11 Feb-10-2004

Terry Oldberg

Engineering, Mechanical Electrical Nuclear Software
Consultant,
USA,
Joined Oct 1999
42
Re: NDT Reliability methodology -- other people interested Dear Wolfgang:

As you have noted, German-English translation of statistical terms can be a barrier to communication about NDT's research protocols. Another, which you have not noted, is that misuse of statistical terms among NDT's native English speakers, native German speakers and others can be a barrier to communication.

For example, in the U.S. Nuclear Regulatory Commission's report on its study of NDT reliability in
pressurized water reactor steam generator tubes, the "probability" of detection cannot be "probability," as this term is used in mathematics. This follows from the fact that probability theory is empirically
invalidated by the test and associated research protocol.

Because "probability" has sometimes been misused
in this way and because this misuse has sometimes obscured flaws in NDT's reliability studies, one should be careful in one's use of this term in discussions such as this one. In particular, I urge avoidance of the terms "Probability of Detection," "POD," "Probability of False Call," "False Call Rate," and similar terms in circumstances in
which the empirical preservation of probability theory has not been demonstrated.

"Sample" is another of these problematic terms. In statistical research, it designates a subset of the
complete set of the objects that are termed "sampling units" in statistics. However, the designers of NDT's
tests have often (almost always, so far as I can determine) overlooked the necessity for specifying the
sampling units. In the wake of this omission, people designing associated research protocols have sometimes obscured the inconsistencies that this causes through redefinitions of what "sample" usually means in statistics.

In particular, a single, arbitrarily selected chunk of material is not an example of what one means by "sample" in statistics. Neither is a set of arbitrarily selected chunks of material. A discontinuity is not necessarily an element of what one means by "sample," though it might be. The term "specs obtained with a sliding probe, especially if you want to calculate the false alarms...
: What is "location with discontinuity" and a "location without" do you calculate with smaples of 1 mm in length, or in samples with length of probe diameter (as the active area) or how do you define a sample location... is it only a test specimen?
: For a good calculation of a false alarm rate you need a high number of samples without discontinuities... a very expensive situation, if you declare only a complete testspecimen to be "one" sample...
: Hope I have put the situation in correct words as I am not a "native english speaking expert in statistics", so some expressions used in german for this topic can't be easily retrived from regular dictionaries (I mostly use http://dict.leo.org - a very good living project of the univ. of Munich), sometimes it is hard for us to hit the definition 100%.
: Conclusion: I like this discussion you did, very much and maybe it gives us some additional ideas to improve our work to the benefit of safety
:
: --
: This message was posted into the NDT.net Forum.
: http://www.ndt.net/wshop/forum/forum-2.htm
: Please do not reply to email: forum@ndt.net
: For your reply to this message go to:
------------ 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|