ML20198E340

Memorandum Raising Concerns Re Statistical Inferences from Comanche Peak Response Team Sampling.Applicant Sampling Studies Design Not Clearly Stated.Comments of Parties Requested by ASLB During CY85.Served on 851112

ML20198E340
Person / Time
Site:	Comanche Peak
Issue date:	11/11/1985
From:	Bloch P Atomic Safety and Licensing Board Panel
To:
References
CON-#485-129 79-430-06-OL, 79-430-6-OL, OL, OL-2, NUDOCS 8511130314
Download: ML20198E340 (5)

Text

,

h

\h \

'h lfD I

l 85-h;Y, LBP~65 gy 12  : 4 Pl2 5 UNITED STATES OF AMERICA i yr NUCLEAR REGULATORY COMMISSION Before Administrative Judges: bhhk-f 31ccb:

4 ,vu.;,

3.k.j r Peter B. Bloch, Chairman "

Dr. Kenneth A. McCollom Dr. Walter H. Jordan Herbert Grossman, Esq.

SERVED NOV1219ss In the Matter of Docket Nos. 50-445-OL & OL-2 50-446-OL & OL-2 TEXASUTILITIESELECTRICCOMPANY,etal.)

-) ASLBP No. 79-430-06 OL (Comanche Peak Steam Electric Station. )

Units 1 and 2) )

) November 11, 1985 MEMORANDUM (Statistical Inferences from CPRT Sampling)

The purpose of this memorandum is to raise Board concerns about the approach to statistical inference apparently being used by the Comanche Peak Response Team (CPRT).1 Although the Staff of the Nuclear Regula-tory Consnission (Staff) is following this issue closely, we are raising our concerns at this time so that they will not become a saurce of delay should evidence gathered by the CPRT be introduced at a later stage of this case.

Our concern is that we have not seen a clear statement of how Applicants have designed their sampling studies. We expect 1

See Transcript of Meeting Between NRC Staff [(Staff)] and Texas Utflities Generating Company [(Applicants)] to Discuss the Official Inspection of Painted Support Welds, September 17, 1985; Transcript of Public Hearing, Plantation Inn, Granbury, Texas, October 2 and 3,1985[betweenApplicantsandStaff].

8511130314 851111 PDR ADOCK 05000445 C PDR 3ys o 2

T J t

Statistical Inference: 2 that evidence from which the Board is asked to draw a statistical inference should include, in effect, the following elements 2

1. Statement of the hypothesis and assumptions,

2. Statement of the level of significance chosen,

3. The test statistic and critical region,

4. Presentation of any computation [or reference to a verified code]

5. A full statement of the conclusions.

The most important reason for adhering to this general formulation is that it clearly displays the relationship between the problem being addressed and the resolution of the problem. For example, if the problem is verifying that wiring has been properly installed in the Comanche Peak Station, then one must decide how to construct the sample of wiring to assure that all the wiring has been properly installed. If everything about wire installation is the same, then it might be proper to treat all the wiring as the same population and take a single sample.

If, however, certain wiring requires far more complex procedures for installation, then it might be necessary to utilize a stratified sample that will permit generalization about the complex procedures looked at by themselves. If the procedures are indeed more complex, then one 2

This particular formulation of requirements is from Wilfrid J.

Dixon and Frank J. Massey, Jr., Introduction to Statistical Analysis; Fourth Edition, McGraw-Hill, New York (1983) at 85.

However, the general principles expressed are believed by the Board to be connon knowledge among statisticians.

4' i

Statistical Inference: 3 would not maF inferences that because simpler tasks have been done  ;

properly that the complex must have been done properly as well.3 Similarly, if it is alleged that certain quality assurance inspec-l tors or certain craft personnel were not qualified, one would seek a stratified sample of their work in order to determine whether each of the questioned employees perfonned adequately. Although the sample might not include the work of every such questioned person (we do not reach a conclusion concerning whether each person's work would need to be included), it would have to be adequate to draw inferences about the work of each questioned inspector. One likely would not draw inferences about a few questioned inspectors by sampling work that was generally done by qualified personnel. There is no basis for generalizing from the work of primarily qua'ified personnel to the work of unqualified personnel.

Special problems may be caused by changes in personnel over time.

For example, it is not clear that one could generalize from work per-formed under Mr. Tolson to work performed under his predecessors, or vice versa. Along the same lines, if there was a time period in which the audit function was neglected, it might not be possible to generalize about that time period from work performed in other time periods.

3 See Tr. (Plantation Inn) at 79. We reach no conclusion about the correctness of Mr. Calvo's point.

.o Statistical Inference: 4

-With respect to the statement of the hypothesis and assumptions (step 1), it is apparent to the Board that a sample must be carefully structured if an inference is to be drawn to rebut the existence of a problem already detected by the Staff or by the Applicants. Generally, unstratified samples are unlikely to prove to be persuasive with respect to specific, identified problems.

A hard question also must be posed about the level of safety that must be assured. With an adequate Appendix B program, there undoubtedly would be some oversights and deficiencies. The project is too large to be perfect. However, if the Board is to be persuaded that there was an inadequate quality assurance program, then a sample must be adequate (given the level of confidence permitted by statistical inference from closely approximate the sample) to / the level of safety contemplated by Appendix B.

If that level of safety cannot be assured, the plant fail to meet Appendix B requirements and to .

would appear togrequire the granting of an exemption from Appendix 8 pursuant to 10 CFR 2.758.

Additionally, to the extent that the FSAR makes promises about the applicability of particular codes, the purpose of the review program is to determine whether those comitments were met. If commitments were not met, then the Station must be corrected or the Staf! tnust determine whether it is appropriate to permit aa amendment of the FSAR.

We seek the comments of the parties concerning the points made in this memorandum, including the points about stratified sampling, cali-bration of the sample concerning allowable levels of safety, the need to relate specific problems (hypotheses) to the sample, and the possible

V 1

Statistical Inference: 5 need for an exemption from Appendix B. Parties may file their responses so that they are received by the Board during calendar year 1985.

FOR THE ATOMIC SAFETY AND LICENSING BOARD I

Peter B. Bloch, Chairman ADMINISTRATIVE JUDGE Bethesda, Maryland