Appendix II Analysis of the Nuclear Regulatory Commission's Probabilistic Risk Assessment for Davis-Besse Report of the Committee to Review the Davis-Besse Nuclear Power Station John C. Lee Department of Nuclear Engineering and Radiological Sciences Ann Arbor, MI 48109 Thomas H. Pigford Department of Nuclear Engineering Berkeley, CA 94720 Gary S. Was Department of Nuclear Engineering and Radiological Sciences University of Michigan Ann Arbor, MI 48109 February 2004 Appendix II Analysis of the Nuclear Regulatory Commission's Probabilistic Risk Assessment for Davis-Besse Report of the Committee to Review the 1. Scope of the Review The U. S. General Accounting Office formed a committee in September-October 2003 to review the oversight that the U. S. Nuclear Regulatory Commission provided on matters related to the pressure vessel head corrosion at the Davis-Besse (DB) Nuclear Power Station. The GAO charge to the committee was to respond to the questions: (1) What probabilistic risk assessment model did NRC use and is it an appropriate model? (2) What was the source of key data used to run NRC's probabilistic risk assessment and were these data valid? (3) What key assumptions implicit in the model did NRC use to govern the estimated risk of different scenarios and were these reasonable? (4) Is probabilistic risk assessment an appropriate tool for making such decision in these instances? (5) How could NRC improve its use of probabilistic risk assessment to make more informed decisions? The committee was initially provided with a set of 53 documents, which included GAO's preliminary analysis of the issues involved and chronology of the DB events during 2001 and 2002. The GAO reports summarized NRC-DB interactions in fall 2001 related to NRC Bulletin 2001-01 on control rod drive mechanism (CRDM) nozzle cracking, the eventual shutdown of the plant on 16 February 2002, and the subsequent discovery of pressure vessel head corrosion. Included also were: (1) Official NRC documents, Generic Letters, Bulletins, and Information Notices transmitted to licensees including Davis-Besse, (2) DB reports submitted to NRC related to the CRDM nozzle issues, (3) NRC documents summarizing the staff's positions and discussions, (4) Summaries of NRC staff presentations to NRC's Advisory Committee on Reactor Safeguards (ACRS) and to the Commission Technical Assistants, (5) Event inquiry report of the NRC Office of Inspector General (OTG) and response from the NRC Chair, (6) Redacted transcripts of OIG interviews of NRC staff, and (7) Transcripts of GAO interviews with NRC staff. The committee reviewed the initial set of documents received from GAO and conducted discussion on the phone and quite frequently via email. One member (GSW) provided a set of initial questions, which GAO used in a meeting with the NRC staff in October 2003. Another member (JCL) met with Mark Reinhart of NRC at the November American Nuclear Society meeting to discuss relevant technical issues and to prepare for a meeting of the review committee with NRC staff, which took place on December 11, 2003. At the meeting, two members (GSW, JCL) discussed technical and management issues with a total of nine NRC officials. Appendix II Analysis of the Nuclear Regulatory Commission's Probabilistic Risk Assessment for Davis-Besse 2 (3) Additional NRC reports on significance assessment of the DB CRDM degradations and the October 2003 OIG review of NRC's oversight on DB, (4) Reports (including one proprietary version) from Electric Power Research Institute and Nuclear Energy Institute, (5) Notes from William Shack, Argonne National Laboratory (ANL), describing his calculation of CRDM nozzle failure probability. (6) DB probabilistic risk assessment (PRA) study performed for NRC by the Idaho National Engineering and Environmental Laboratory, (7) Transcripts of several ACRS meetings during 2001-2003, and (8) Select papers in engineering journals and proceedings. The committee conducted an extensive review and discussion on the probabilistic risk calculations performed both by the FirstEnergy Nuclear Operating Company (FENOC) and NRC for Davis-Besse. One committee member (JCL) also developed a simplified analytical model to determine the CRDM failure probability, which provided a rough check on numerical calculations performed at ANL Following the 11 December 2003 meeting with the NRC staff, the committee made an effort to follow up on a number of questions that required additional information or clarifications. One essential piece of information is the core damage probability due to the postulated CRDM failure and ejection that NRC actually used in connection with the decision to allow continued DB operation until February 16, 2002. After a long wait, finally on February 24, 2004, the committee received a response from Jin Chung, Richard Barrett, and Gary Holahan, summarizing, to the extent they could reconstruct, how NRC arrived at key quantitative risk estimates in November 2001. We present in Section 2 key findings of the committee on NRC's oversight related to the DB issues. We provide responses to the first four GAO charges in Sections 3 through 6, in a slightly restructured format, covering (a) PRA methodology and data used in NRC's risk assessment, (b) assumptions and uncertainties in the risk assessment, (c) relevant regulations and guidelines, and (d) November 2001 NRC decision. Our response to the fifth GAO charge is finally presented in Section 7. 2. Key Findings of the Committee The committee presents key findings of its review on NRC's oversight on Davis-Besse and related safety and regulatory issues: (1) NRCs Risk Analysis for Davis-Benne (a) To guide a risk-informed decision on whether to grant an extension beyond its December 31, 2001 date for shutdown of Davis-Besse for nozzle inspection, NRC relied on its PRA of risks from crack-induced failure of control-rod housing nozzles. The calculated risk was incorrectly small because the calculations did not consider corrosion of the reactor vessel due to boric acid in coolant leaking through the cracks. The calculated risk was also subject to large uncertainties. As a result, NRC staff found it difficult to balance results of quantitative risk calculations against qualitative considerations. Regulatory Guide 1.174 provided little help in this regard. (b) NRC did not perform uncertainty analysis in applying PRA in the DB decisionmaking process and there was confusion regarding the interpretation of core damage frequency (CDF) and core damage probability (CDP) as risk attributes within the framework of RG 1.174. NRC staff should have recognized large uncertainties associated with the CDF estimated for CRDM nozzle failures |