ML20141K014

Summary of 851111 Meeting W/F Webster (J Benjamin & Assoc) in Mountain View,Ca Re Statistical Sampling Approaches in App D of Applicant Program

ML20141K014
Person / Time
Site:	Comanche Peak
Issue date:	01/02/1986
From:	Lurie D NRC
To:	Calvo J Office of Nuclear Reactor Regulation
References
NUDOCS 8601220113
Download: ML20141K014 (1)

Text

t)egg nebed cymc6 gs " "% UNITED STATss r

u d

, flUCL3AR REGULATORY CcMMISSION g, gg gg 3 //ASHINGTON. D. C. 20553

[hh*f,?

% a-iM 2 iS6*o NON~FOR: J. A. Calvo, NRR/ICSB FROM: D. Lurie, RM/BMI

SUBJECT:

MEETING WITH FRED WEBSTER OF J. BENJAMIN & ASSOCIATES On November 11, 1985 I met with Dr. Webster to discuss the statistical sampling plan for Comanche Peak. The meeting took place at J. Benjamin &

Associates' office in Mountain View, California. Dr. Webster and I were the only participants at the meeting.

The discussion focused on the two sampling approaches given in Appendix D, "CPRT Sampling Approach Applications and Guidances" of the CPRT prcgram.

The objective of the sampling plan was to provide a 95/5 assurance for each

" population" (work process) in the plant. Appendix 0 identified two approaches to sampling a population. The first approach was denoted as the " classical" dpproach, and the second approach as the "Baysian" anproach. Of the two approaches, the cla:sical approach is somewhat more wonservative. Dr. Webster elected to use the Baysian approach. From a strict probabilistic considerations, the Baysian approach, gives a 93/5 assurance at best.

There are several points that can be raised in connection with the sampling plan at Comanche Peak.

- The choice of 95% level of assurance is arbitrary. Perhaps a 90%

asstrance would have been sufficient. It is not the statistics that makes the plant safe. The statistics is only a way to quantify the assurance. In light of other considerations, this quantification provides only a lower bound for the true assurance.

,/

The sampling plan supplements many other inspections and quality control programs that have been implemented at Cemanche Peak. Any QA activity performed in parallel with the sampling process clearly enhances the stated assurance, even if this enhancement cannot be quantified.

The assurance increases with the sampling size, provided no additional defectives are found. Cases did arise where the sample size increased when items for one population were examined which shared attributes with another population already sampled.

The sampling assurance is enhanced when defective items are repaired or replaced. For small populations this enhancement can be significant.

Items of importance were subject to 100% inspection, yielding 100%

assurance that no items is defective.

When a defective item was located, it served to identify a potential error path, which lad to further sampling. This activity increased the effective sample size, generally giving a still greater assurance.

860122011396g0 g cc. C. Trammell, NRR /

R. Hartfield, RM PDR A

ADOCK O PDR l/?

F

{p