Search Options | ||||
Index | Site Map | FAQ | Facility Info | Reading Rm | New | Help | Glossary | Contact Us |
July 23, 2002 The Honorable Richard A. Meserve Dear Chairman Meserve:
During the 494th meeting of the Advisory Committee on Reactor Safeguards, July 10-12, 2002, we reviewed the draft final Revision 1 to Regulatory Guide (RG) 1.174, "An Approach for Using Probabilistic Risk Assessment in Risk-Informed Decisions on Plant-Specific Changes to the Licensing Basis," and to the Standard Review Plan (SRP), Chapter 19, "Use of Probabilistic Risk Assessment in Plant-Specific, Risk-Informed Decisionmaking: General Guidance." During this review, we had the benefit of discussions with representatives of the NRC staff. We also had the benefit of the documents referenced. RECOMMENDATIONS
DISCUSSION The publication of RG 1.174 and associated SRP Chapter 19 in 1998 was a major milestone in the NRC initiative to risk inform the regulations. RG 1.174 introduced the concept of an integrated decisionmaking process that had as inputs risk information, considerations of defense in depth, and sufficient safety margins. The Guide also defined acceptable ranges of values for the possible increases in core damage frequency (CDF) and large early release frequency (LERF) that could result from proposed changes in the licensing basis. The approach to the use of risk information established in RG 1.174 was consistent with the philosophy of a risk-informed, rather than risk-based, regulatory system. As such, the Guide did not determine acceptance in terms of strict numerical values for DCDF and DLERF and did not specify how to integrate various inputs to the decisionmaking process. Also, the Guide stated that the scope, level of detail, and technical acceptability of the PRA should be commensurate with the application. Because the explicit use of risk information from PRAs for regulatory purposes was relatively new in the late 1990s, the staff recognized in SRP Chapter 19 that licensees probably would not have a PRA that analyzed all significant plant modes and accident initiators. It was also recognized that the quality of available PRAs varied widely. Because risk information was to be utilized in an integrated decisionmaking process, it was accepted that such incomplete PRAs could still provide useful insights into the risk impact of proposed licensing changes, thus leading to more effective regulatory decisions. Although this approach has been successful in the development of some risk-informed licensing changes, such as risk-informed inspection programs, it had the unfortunate consequence that it did not encourage the development of full-scope PRAs for all operational modes nor the use of rigorous methods in developing risk information. Incomplete and less-than-rigorous PRAs undermine the credibility of the entire risk-informed regulatory process. We believe that the slow progress of risk-informed initiatives can be attributed, in part, to this lack of credibility. The proposed revisions make no substantive changes to the existing RG 1.174, and publication of these revisions may send the wrong message that incomplete PRAs are acceptable for a broad range of risk-informed changes to the licensing basis. Therefore, these revisions should not be issued. Revision to RG 1.174 and the associated SRP Chapter 19 should reflect the progress of the last five years, as exemplified by PRA standards that provide new consensus guidance for a high-quality PRA and the industry peer review process. In view of these developments, the revised guidance should specify that a licensee who wishes to take advantage of risk information produce such information using methods consistent with rigorous consensus standards and include all significant sources of risk from all plant modes. Such a provision would bolster the credibility of risk-informed decisions and reduce the staff and licensee effort required to assess risk implications of licensing changes. The PRAs should include rigorous uncertainty analyses, at least for parameter uncertainties. This would allow more focused attention on those sources of uncertainty that are much more difficult to address, such as model uncertainty and incompleteness. When approximations (e.g., bounding estimates of the risk due to plant modes not rigorously analyzed) or approximate methods are used, they should be justified. Sensitivity analyses should be used judiciously and only after an uncertainty analysis has been performed. We look forward to working with the staff on these important issues. Additional comments by ACRS Members Dana A. Powers and John D. Sieber
are provided below.
Additional Comments by ACRS Members Dana A. Powers and John D. Sieber Our colleagues have the laudable desire to encourage improvements in
the scope and depth of probabilistic risk assessments being utilized by
licensees especially when they seek risk-informed changes to licensing
bases. Our colleagues are, however, seeking improvements with such demands
for rigor and stringency, regardless of need, that they may alienate some
applicants. Such demands will increase the burdens associated with RG
1.174 and SRP References:
|