ML20045B205

From kanterella
Jump to navigation Jump to search
Discusses H Lewis Position on Diesel Generator Reliability. Two Options Suggested by H Lewis Delineated.Threshold Criteria Not Supported by Statistics Involved
ML20045B205
Person / Time
Issue date: 03/18/1991
From: Fraley R
Advisory Committee on Reactor Safeguards
To: Beckjord E
NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES)
Shared Package
ML20042D089 List:
References
FRN-57FR14514, REF-GTECI-B-56, REF-GTECI-EL, RULE-PR-50, TASK-B-56, TASK-OR AE06-1-055, AE6-1-55, NUDOCS 9306160411
Download: ML20045B205 (1)


Text

_ _ _ _ _

Afo f -/

f~M6 tecLQ

~-

~

'37 u

w. k oJJ e ca

/

s'o

g. W UNITED STATES y

~,,

NUCLEAR REGULATORY COMMISSION 5

E ADVISORY COMMITTEE ON REACTOR SAFEGUARDS

,f WASHINGTON, D. C. 20555 March 18, 1991 w

T b

g MEMORANDUM FOR:

Eric S. Beckjord, Director Office of Nuclear Regulatory Resear,ch FROM:

R. T raley Executive Dire tor

SUBJECT:

DIESEL GENERATOR RELIABILITY

Reference:

Letter dated February 19, 1991, from H.

W.

Lewis, ACRS, to J.

W.

Johnson, RES,

Subject:

Confidence Limits Hal Lewis has just informed me that you sent him a letter that suggests the differences between him and the staff on diesels are matters of personal taste rather than technical substance.

He assures me that this is not the case and maintains that the choice of limits by the staff results in too large a probability of enclosing the correct value of the reliability and gives too large a reliability at the upper end of the interval.

A factor of ten in the bounds of the confidence interval is inappropriate fo'r regulation.

The upshot is that the threshold criteria is not supported by the statistics involved.

He has provided the following table (using the poisson surrogate for the binomial distribution and Johnson's definition of a confidence interva1).

Under1vina Failure Rate Fraction of Measure-ments that Enclose It 0.1 0.9953 0.2 0.9825 0.4 0.9921 0.8 0.9526 1.6 0.9763 l

l Dr. Lewis has suggested thct there are two options - either do the Bayesian statistics correctly (which i.ll show there is little information that can be derived f:

such small numbers of failures) or declare the threshold to be arbitrary which he has favored from the beginning.

He has asked that I inform you of his position.

9306160411 930422 57 14514 P

(j)