ML20035D755
| ML20035D755 | |
| Person / Time | |
|---|---|
| Issue date: | 10/01/1991 |
| From: | Taylor J NRC OFFICE OF THE EXECUTIVE DIRECTOR FOR OPERATIONS (EDO) |
| To: | Ward D Advisory Committee on Reactor Safeguards |
| Shared Package | |
| ML20035D698 | List: |
| References | |
| ACRS-2791, NUDOCS 9304140041 | |
| Download: ML20035D755 (3) | |
Text
__
o
, s.* %
[D v t Y h UNITED STATES i
NUCLE AR REGULATORY COMMISSION gg y
h ASHINGTON. D. C. 20555 e
N'...+/
001 J l Li l
Mr. David A. Ward, Chairman Advisory Comittee on Reactor Safeguards U.S. Nuclear Regulatory Commission Washington, D.C.
20555
$i
Dear Mr. Ward:
Your letter of July 19, 1991, to the Chairman identified concerns of the ACRS about the consistent use of probabilistic risk assessment by the NRC staff.
The letter notes a number of symptoms of what you identify as more general problems, the lack of a coherent policy on the use of PRA by the staff, and i
the limited number of staff experts in PRA.
With respect to the general problems you have identified, we have the following observations:
The level of sophistication of staff PRAs and of internal and external o
reviews clearly varies among applications. This is entirely appropriate, recognizing that certain analyses require more detailed consideration than others and that the underlying technical issues vary in potential safety significance. For example, more PRA resources have been assigned to reviews of advanced reactor design submittals and to certain key generic issues (e.g., station blackout) than to the analysis of other generic issues. Further, the lack of adequate data has influenced our use of PRA in many applications.
The policy of the staff is to make best estimates of risks and costs.
o Where possible, this means using realistic assumptions and statistical However, in many analyses the data are sparse and the best that means.
can be done are point estimates.
In other cases, no data are available and only conservative estimates are possible. Where such departures are taken from the ideal, they are to be clearly displayed and their effect on the proposed decision explained, The level of sophistication of uncertainty analysis also varies among o
applications of PRA. Such variation is appropriate because of variations in the significance of safety issues and the availability of supporting data. While the recent completion of NUREG-1150 has made the issue of uncertainty analysis much more apparent, the staff has not completely studied its implications for a more general implementation in j
the regulatory process. One important aspect of such a study would be the development of improved methods for generating the data needed in such analyses.
ATTACHMENT D l
9304140041 920312
/[)
David A. Ward, Chairman 2
The staf f is well aware of the uncertainty and unreliability of PRA o
results, particularly reliance on bottom line numbers. Assessing this uncertainty and unreliability is an important part of any decision.
However, I don't believe that it is practical at this time to employ a decision-making algorithm with a prescribed confidence level as you suggest. With a few possible exceptions the data required to implement such a method is simply not available.
Substantial effort is being expended to improve the PRA knowledge of the o
NRC staff.
During FY 92, a total of 20 4-day duration PRA courses ranging froin introductory to advanced level are scheduled for presentation to appropriate staff.
In addition, NRC management has recognized a need for a " cultural shift" in staff understanding of risk and PRA considerations. Beginning in 1990, the NRC Technical Training Center staff were provided PRA training and during 1991 have subsequently revised the reactor technology lesson plans to specifically address risk perspectives as each topic or system is discussed. The objective is to inculcate risk perspectives and appropriate application of PRA insights into NRC personnel as they proceed through the technical training program.
The staff has noted previously (e.g., in SECY 91-161 [on advanced o
reactor review schedules]) that available personnel with the requisite backgrounds in probabilistic analysis and accident phenomenology are at a premium.
Efforts have been made to expand the staff's capabilities, with moderate success; these efforts are continuing.
The staff's ability to recruit PRA experts and persons with the potential to become PRA experts has been hampered by competition with other federal agencies and private industry.
In this circumstance, the available staff resources must be carefully prioritized to optimize their influence.
In many cases, contractors have been used to supplement the staff, with varying degrees of success.
We agree that the issue of obtaining additional staff expertise is not o
limited to the PRA field. The staff is now working to recruit, for example, people with expertise in digital instrumentation and control systems.
Thus, while it does not appear that major problems now exist in the use of probabilistic risk assessment by the staff, I believe that a review of the staff's PRA activities is appropriate. This review will consider what additional guidance to the staff would assure the consistent development, content and use of PRA within the NRC. Since all of the program offices have an interest in the application of PRA, an interoffice group will be established to conduct such a review.
I would expect that a review could be lE
//
L
1 s
Mr., David A. Ward, Chairman 3
completed in a few months and will keep the Committee informed of its work and findings.
l Sincerely,
/
\\
W /~
es M. Tpflor xecutive Director for Operations cc:
The Chairman Comissioner Rogers j
Commissioner Curtiss Comissioner Remick SECY l
4 i
i 20
'