ML20094P606
| ML20094P606 | |
| Person / Time | |
|---|---|
| Site: | Byron  |
| Issue date: | 08/14/1984 |
| From: | Judson V BUSINESS & PROFESSIONAL PEOPLE FOR THE PUBLIC INTERES |
| To: | Callihan A, Cole R, James Smith Atomic Safety and Licensing Board Panel |
| References | |
| OL, NUDOCS 8408170298 | |
| Download: ML20094P606 (4) | |
Text
050r 3
BPI Business and Professional People for the Public Interest Telephone: (312) 641-5570 j
Chicago, Illinois 60602
(
109 North
Dearborn Street,
Suite 1300 e
e August 14, 1984
%MED Ivan W. Smith, Esq.
Dr. Richard F. Cole'84 AG916 P1 :13 Administrative Judge and Administrate Judge Chairman Atomic Safety and Atomic Safety and Licensing Licensing Board k~O NiiC h Board U.S. Nuclear Regulatory 9.R;.Ncs U.S. Nuclear Regulatory Commission Commission Washington, D. C. 20555 Washington, D. C.
20555 Dr. A. Dixon Callihan Administrative Judge Atomic Safety and Licensing Board c/o Union Carbide Corporation P. O. Box Y Oak Ridge, Tennessee 37830 Re:
In the Matter of Commonwealth Edison Company (Byron Nuclear Power Station, Units 1 and 2)
Docket Nos. 50-454 and 50-455 de Gentlemen:
The written testimony of Dr. Eugene P. Ericksen that was sent to you yesterday contained typos on pages 6,
11 and 17.
Please substitute the enclosed corrected pages.
Very truly yours,
\\ 0 3-\\.Cu o
Victoria A. Jud'slonwn l
V VAJ:mam
- encls, cc: Service List D6 rectors James M. Alter Staunton O. Flanders Byron S. Miller Staff James W Asniey Lucy 8. Ascon Hebet 8. Fned Elena 8. Mulcahy Asenander Pohkoff Juhe M. Kuzera
,,g,,,0# j. Bowman PressNot b'"
O"## *#
Ronal ski A en koff Julian L 8erman Douglass W. Casset, Jr.
Martin C. Hausman Jeremy Warburg Russo Douglass W. Casset, Jr.
Jeanne L Yeidel Rschard P. Kiphart Frank Cacero Chnstie Hefner Alan Saks General Counsel Admensstrative A,s,stant v,c7$e**!ents
"."o'r$E8'"
7"d 8 K nin S*NIUu't ir "
?a"," Sa"**,"*j'ames
'"'*"'d*"
'u E
$E C"r'"e** "e'r"'
'"'O'"AN" J'
Luts E. 06az. Pere'z Robert J. Vo'llen H
aw O' U" "
Robert 8. Latton Jane M. Whscher r la eve F f er E.
Mc' 11 n we man Leon D. Finney John L McKnight Wayne W. Whalen N
]
?
8408170298 840814 PDR ADOCK 05000454
, D LS J T
y
_ 'want to be assured that sample ' sizes 'were sufficiently o
^
- largeLto be confident of the'resul's..This would-have-t enabled i the R'einspection Program to ^ establish acceptable Leonfidence levels and reliabilities based -on.the importance of the element..- Confidence levels indicate how certain'a
-statistician.is-that his or her results are correct.
~
Reliabilities reflect the percentageof inspections which are correct. For inspection elements where the risks caused by a poor quality are great, we might want-to be certain that all were correct-and, therefore, reinspect all ele-ments.
For inspection elements where the risks are not as great, but still substantial, we might want to be quite sure that-99.9 percent were correct.
For other inspection elements which are less safety significant, we might be satisfied if we were reasonably certain that 99 percent were correct.
In order to determine the amount of certain-ty and perfection required for each element, choices should have been made using engineering judgments.
These judg-ments, along with their rationales, should have been deter-mined when establishing the program and clearly stated in the reinspection report.
A reasonable reinspection program might have required the following reliabilities and confi-dence levels for the following types of elements.
Type of Element Reliability Confidence Level Critical to safety 100%
100%
Very important'to safety 99 9%
99%
-Somewhat important to safety 99%
95%
Least important to safety 90%
95%
6
("'
Assumption (2) was' violated because inspectors were not
-homogeneous; different inspectors had different probabili-ties of success.
Assumption (4) was violated because in-spections were not randomly chosen; the selections of inspec-tions were not independent from each other.
Q16: 'What is the basis for your conclusion that inspectors were not homogeneous?
A16:
Where inspectors are not homogeneous there will be simi-larities between inspections made by the same inspector.
This creates a commonality within the cluster which can be measured by the "intraclass corre la t i o n."
The intraclass correlation can range from a value slightly less than zero to + 1.0.
If the intraclass correlation is equal to zero, it means that inspectors are homogeneous and there is no increase in variance associated with cluster sampling.
If the interclass correlation is greater than zero, then inspectors are not homogeneous.
We can une data f rom Appendix B of the Reinspection Report to compute intraclass correlations.
The computa-tions show that for Hatfield, Hunter and Pittsburgh Testing Laboratory, each contractor's overall intraclass correla-tion was greater than zero.
These positive intraclass
~
correlations indicate that inspectors were not homogeneous.
Another indication of the lack of homogeneity among inspectors is seen from the results of "F tests."
The F test is a common statistical tool that can be used to 11
p j
r Third,' Edison used an inappropriate formula in calcu-lating reliabilities.
Two assumptions of the formula were d
violated: inspections were not randomly selected and inspec -
. tors were:not homogeneous.
Fourth, Edison did not account for the added uncertainty created by clustering of inspections by inspector.
For these reasons, the sampling design of the Reinspec-I tion Program and the statistical analysis'of the Reinspec-tion Report are inadequate to support Edison's general I
conclusions about work quality and inspector qualifications.
i L
I e
i i
(
~
s I
17 l
t