ML20217E660

Comment on Draft NUREG-1602 & Draft RGs DG-1061,DG-1062, DG-1065 & Draft Std Review Plan Chapters 3.9.7,16.1 & 19, Addressing Use of PRA in Regulatory Process

ML20217E660
Person / Time
Site:	Calvert Cliffs
Issue date:	09/30/1997
From:	Cruse C BALTIMORE GAS & ELECTRIC CO.
To:	NRC OFFICE OF ADMINISTRATION (ADM)
References
FRN-62FR34321, RTR-NUREG-1602, RTR-REGGD-XX.XXX, TASK-*****, TASK-RE 62FR34321-00020, 62FR34321-20, NUDOCS 9710070120
Download: ML20217E660 (11)

Text

. .

Cucus II. Cat >sr. n Baltimore Gas and Electric Company V ce President REnElVED C*en Cm suaear Pewer nani Nuclear Eneigy 1650 Calvert Cliffs Parkway 1997 OCI -2 PH 3: 06 '"

• Y # "')' "

• 57 pp 410 495-4455 U (U MLu,p9/p m RULES & DR LPMCH US NRC f .2 f /2 j f f _2 /

0 [ j f' c /) e t # O September 30,1997 g

U. S. Nuclear Regulatory Commission Washington, DC 20555 ATTENTION: Chief, Rules and Directives Branch

SUBJECT:

Calvert Cliffs Nuclear Power Plant l Unit Nos.1 & 2; Docket Nos. 50-317 & 50-318 l Comments on Draft NUREG 1602, Draft Regulatory Guides DG 1061, j

' DG-1062, DG 1065, and Draft Standard Review Plan Chapters 3.9.7,16.1, and

19. Addressing Use of Probabilistic Risk Analysis in the Regulatory Process

REFERENCE:

(a) Federal Register Notice, dated June 25, 1997, Use of PRA in Plant Specific Reactor Regulatory Activities: Proposed Regulatory Guides, Standard Review Plan Sections, and Supporting NUREG (62 FR 34321)

Baltimore Gas and Electric Company is a strong supporter of the effective use of Probabilistic Risk Analysis (PRA) to aid in the safe and cost effective operation of nuclear power plants. We, therefore, are pleased that the Nuclear Regulatory Commission (NRC) is pursuing risk-informed regulation and for the opportunity to provide comments.

However, we are concerned that without significant changes to the draft regulatory guidelines, risk-informed regulation is in jeopardy. It appears that these documents, as currently written, will not reduce the regulatory uncertainty in this emerging area. The associated uncertainty in effort and cost, and the j limited allowable benefits will likely discourage future PRA applications. Discouraging applications is inconsistent with the Commission's PRA policy statement to increase the use of PRA technology regulatory matters and to reduce unnecessary conservatism associated with regulatory requirements.

Specifically, the requirement of an absolute numerical criterion, the need for uncertainty evaluations, the lack of clarity associated with the quality required for the PRA, and the apparent inconsistency between the requirements for the pilot applications and that stated in the draft documents are of concern.

It appears to us that a framework which encourages growth in both the application and quality of PRAs has not been achieved. Baltimore Gas and Electric Company's comments are, therefore, focused on issues that we believe will help to reconcile these concerns.

9710070120 970930 2 PDR , ,

I i

. Chief, Rules and Directives Br::nch Sept:mber 30,1997

,- Page 2 We also believe, due to the magnitude and importance of the draft documents, that adequate time to perform an cH'ective review was not made available. Therefore, the lack of comments in a specific area does not indicate agreennent.

Should you have questions regarding this matter, we will be pleased to discuss them with you.

Very truly yours, 1.

t f CilC/SJR/bjd

' , , w-

Attachment:

(1) Comments on Draft NUREG 1602, Draft Regulatory Guides DG-1061, DG 1062, DG-1065, and Drah SRP Chapters 3.9.7,16.1, and 19, Addressing- Use of

.Probabilistic Risk Analysis in the Regulatory Process cc: Document Control Desk, NRC H. J. Miller, NRC R, S. Fleishman, Esquire Resident inspector, NRC J. E. Silberg, Esquire R.1. McLean, DNR Director, Project Directorate 1-1, NRC J. H. Walter, PSC A. W. Dromerick, NRC

A'ITACllMENT (1) >

1 l

Comments on Draft NUREG-1602, Draft Regulatory Guides DG-1061, DG-1062, DG-1065, and Draft Standard Review Plan Chapters 3.9.7, 16.1, and 19, Addressing Use of Probabilistic Risk Analysis in the Regulatory Process 11altimore Gas and Electric Company Calvert Cliffs Nuclear Power Plant, Units 1 and ,1, September 30,1997 )

ATTACilMENT (1)

COMMENTS ON DRAIT NUREG 1602, DRAIT REGULATORY GUIDES DG 1961. DG 1962, DG 1065, AND DRAIT STANDARD REVIEW PLAN CilAPTERS 3.9.7, 16.1, AND 19. ADDRESSING USE OF PROPAHILISTIC RISK ANALYSIS IN THE REGULATORY PROCESS NUMERICAL ACCEI'TANCE GUIDELINES

'the following comments are in reference to DG-1061, Section 2.4.2.1, page 14.

The current industry Probabilistic Risk Analyses (PRAs) do not support an upper level absolute acceptance guideline (a mean core damage frequency at or about IE 4 or a mean large early release frequency [LERF) at or about lE 5). Completeness, conservatism and non consuvatism, and the resulting inconsistencies between PRAs are issue that will take time to resolve. In addition, the compiexity and dynamic nature (due to data, procedure changes, plant modifications, external hazard variations, etc.) make the dependency on an absolute number unmanageable. Not only will such an approach be difficult to manage and regulate, it will shift the industry's focus from discovering risk insights to achieving low calculated core damage frequencies (CDFs). It will also result in penalizing those plants with high calculated CDFs, even if the high CDF is solely due to perfonning a quality PRA.

l In addition, there is also a configuration issue associated with the potential that plant changes made l outside the risk informed regulation framework that meet current licensing practices could result in uncalculated risk increases. These risk increases could change previously approved submittals and result in a complex con 0guration control system. We, therefore, believe the only numerical guidelines should be based on the change in risk. A general acceptance guideline of IE 6 is recommended.

Baltimore Gas and Electric Company (DGE) also believes that it is appropriate to pursue a long-term objective of conforming PRAs to standards that result in the ability to compare the risk between plants with confidence that the calculated differences relicct real risk differences. Such an object.ve should be clearly stated and an appropriate means and target date developed (i.e.,2005),

in addition, since comparing the risks of all plants obviously cannot be achieved without considerable cost, the improved benefit ofichieving this objective r:ecds to be defined. The defined benefits should include an elketive risk informed means to achieve relief from low risk significant regulatcry compliance issues. By allowing real near term operating issues to be addressed, the vahie of having a quality PRA will not be in question.

If a framework that is not dependent on an absolute CDF and LERF number is not a viable option, then the guideline should be dependent on the quantitative health objectives and not the surrogate values of CDF: lE-4 and LERF: lE 5. Please note however, that we believe such a choice will be adverse to achieving quality industry PRAs and adverse to promuting PRA applications.

OUAl.ITY ASSURANCE

l. In reference to DG 1061, Section 2.7. page 22, the following five bullets are listed identifying the quality requirements for a PRA application, e Utill:e persormel quahpedfor the analysis, it was stated at the workshop that quallfied personnel referred more to the use of experienced staff, as opposed to the establishment of a formal qualification program. Baltimore Gas and Electric Company believes that a formal program should be expected. Clarification of the expectation for qualified staff should be included in DG 1061-1

. J

ATTACllMENT (1)

COMMENTS ON DRAFT NUREG 1602, DRAIT REGULATORY GUIDES DG 1061 DG 1062. DG 1965 AND DRAIT STANDARD REVIEW PLAN CIIAPTERS 5.9.7 16.1 AND 19 ADDRESSING USE Ol' PROHAHILISTIC RISK ANALYSIS

. IN Tile REGULATORY PROCESS e

Utilize procedu es that ensure control ofdotwnentation, including revisions, andprovidefor the independent review. verification or checking ofcalculations and information used in the analysis (an independent peer can be used as an important element in this process).

The parenthetical implies a peer review may satisfy, at least in pan, the independent review requirement. Is it the intent to say that the peer review satisfies this requirement only if such a review performs detailed checking of calculations in the analysis? %c guidance should be revised to clarify the NRC's position prior to final issuance.

• Provide documentation and maintain records in accordance with the guidelines tre Ree' ion 3 of this guide.

The documentation required in Section 3 appears to be inconsistent with that required in the pilot applications. An application like emergency diesel generator allowed outage time extension requires nearly all aspects of a pRA in order to assess the risk. Therefore, the documentation requirements listed on page 25 would result in a submittal far greater than that of the Individual Plant Examination requiring nearly all the tier two PRA documents. At Calvert Clifts Nuclear Power Plant, this submittal would be well over 50 volumes.

If the intent of this section is to " illustrate the scope and quality of the engineering analysis,"

then the infoimation requested should be stated as being summary in nature to the extent necessary to provide confidence in the application.

Baltimore Gas and Electric Company recommends the following approach: In lieu of providing the large amount of data requested, the application should include a summary of the differences of the utilities' PRA from that of NUREG 1602. This summary should also include a discussion of why each difference is acceptable for the application being pursued. Such an approach will result in a self-assessment of the quality of the I-RA by the licensee prior to the submittal of an application. As such, the regulator is provided a means to cost effectively assess the quality of the PRA used for each application and the licensee is protected from future enforcement that could result from unclear quality expectations. This simple and cost effective approach removes considerable regulatory uncertainty, it also requires the role of NUREG 1602 to change from defining a state of-the art PRA to defining the quality of a PRA necessary to meet the regulator's

[ expectations. Note that this change also increases the importance of NUREG 1602 and necessitates the removal of some unreasonable expectations. See NUREG 1602 comments.

Finally, it is recommended that this bullet be reworded to " maintain records In accordance with the guidelines in Section 3 of this gulde" .ilnce providing documentation is a reporting requirement, not a quality requirement.

• Providefor a : Independent auditfunction to verify quality (an independent peer review can be usedfor thispurpose).

It is assumed that the quality audit function described here is an audit of the PRA >rogram and not each application. The guidance should be revised to clarify the NRC's position prior to final issuance.

2

NITACllMENT (1)

COMMENTS ON DRAFT NUREG.1602 DRAIT REGULATORY GUIDES DG 1061, DG 1062. DG.1065 AND DRAFT STANDARD REVIEW PLAN CilAPTERS 3.9.7, 16.1. AND 19, ADDRESSING USE OF PROllAlllLISTIC RISK ANALYSIS IN Tile REGULATORY PROCESS In addition, the peer review objectives and scope described in NUREG 1602 appear to be very onerous. Since there is no other peer review description provided, please clarify the scope of the peer review expected to satisfy DG 1961. If an independent review is performed in accordance with llullet 2, then can it be assumed that the peer review that would satisfy this requirement is a high level check of assumptions, methodology, and conclusions?

e Utill:e procedures that ensure appropriate attention and corrective actions are taken ff analysis or information usedin previous decision making is determined to be in error.

Due to the dynamic nature of PRAs (periodic updates due to data, plant configuration, analysis changes), it is unclear as to what degree of change in the PRA results would warrant a re-submittal of previously approved applications. Since this liability grows with each submittal made, it is important to understand when a small change is not considered significant !!altimore Gas and Electric Company recommends that such an update would only occur when there is a significant error that invalidates the conclusions of a previous submittal. The guidance should be revised to clarify the NRC's position prior to final issuance.

2. It is 11GE's understanding that the 1".A quality requirements for the risk informed Technical l

Specification change were satisfied through the use of a cross-comparison of key issues and results of several PRAs associated with the pilot application. It is unclear on review of DG 1061 as to whether this practice would be acceptable in the future, and if not, then the pilot application did not meet the quality expectations. The guidance should be revised to clarify the NRC's position prior to final issuance.

Tile ROLE OF NUREG-1602 In reference to page til,last paragraph,it appears that NUREG 1602 is not adequate for the role that is needed it is stated that the most demanding applications define the level of technleal detail contained in the NUREG It was also stated at the August 1997 workshop that the definition of the minimal requirements for a particular application was going to be identified through the application approval process. This approach leaves considerable regulation uncertainty, it fails to provide enough up front guidance to allow any degree of confidence that pursuing a risk informed application can be done in a cost effective manner, it also adds an enormous regulatory risk factor in that a risk informed application that was done in good faith could be later inspected to the only PRA quality guidance available (the technical requirements for the most demanding applications). Such an inspection would likely be frustrating and costly.

The lix is to design NUREG 1602 as a document that defines the quality of a PRA necessary to meet the regulator's expectations for a good PRA. As stated in the discussion of quality above, NUREG 1602 should be used as a benchmark document where differences are summarized in each submittal. This fix implies that NUREG 1602 should be reasonable in its expectatiora and consistent with current good practices. State-of the art or best practices could still be included, but they should be clearly identified.

The below comments are made with this position in mind and attempt to identify the major issues in addition to unclear statements.

3

< J

NITACllMENT (1)

COMMENTS ON DRAIT NUREG 1602, DRAFT REGULATORY GUIDES DG.1061, DG 1062, DG-1965 AND DRAIT STANDARD REVIEW PLAN CilAPTERS 3.9.7, 16.1, AND 19, ADDRESSING USE OF PROHAHILISTIC RISK ANALYSIS IN THE REGULATORY PROCESS Initlnting Events (Ssstion 2.1.1)

1. In Section 2.1.1.1, page 2 3, the inclusion of" unplanned controlled shutdown" events appears to be inappropriate. Such events will not challenge the Reactor Protective System, will have considerably longer human action responses for many actions, and are likely to have different system success criteria due to lower decay heat. Haltimore Gas and Electric Company recommends that this initiator type be removed.

2. It is unclear as to what is meant by an initiating event that has occurred during shutdown conditions.

This reference should be clarined or removed.

Data Analysis (Section 2.1.4)

1. In Section 2.1.4.1, page 2 21. Equipment Reliability, it is stated: "i'or test surveillance or other demands for which actual run times are distinctly less than the length of the mission time modeled in the PRA, it should be determined whether the failure rate derived for the truncated tests or demands is applicable over the mission time." While we agree that the obvious differences between test and actual demands need to be addressed, exclusion of routine surveillance test in reliability calculations appears inappropriate, if such an approach is used, there will be little plant-=pecine data included in the reliability calculations for standby equipment.

2. In Section 2.1.4.1, page 2 22, Common Cause Failures, it is stated: "Since there is generally insufficient data to derive plant specific estimates of the common cause failure parameters, generic data should be used." This statement is too simple and could be misleading. As stated in Electric Power Research Institute document TR 100382, A Database of Common Cause Events for Risk and Reliability Applications, it is " essential to reinterpret the (generic) event in light of the specific characteristics of the system under consideration." Electric Power Research Institute document TR 1002747, Common Cause Data Analysis Tool (CCDAT) User's Manual, provides a tool to help in the development of plant specific, common-cause failure frequency parameters, in addition, it is BGE's opinion that it is also valuable to update generic data with plant specine data. This statement should be rewritten to state: "Since there is generally insufficient data to derive put-specific estimates of the common cause failure parameters, generic datau be used as long as it is appropriately mapped to the specific characteristics of the system under consideration."

Iluman Actions (Sestion 2.1.5)

1. In Section 2.1.5.1, page 2 24, of NUREG 1602 states: ". . , credit for recovery uctions may not be given unless at least some procedural guidance is provided or operators receive frequent training that would lead them to perform the required actions." Although most credited human actions do have procedural guidance, some of the actions rely on the operators' expertise and their overall knowledge cf the plant. In Calvert Cliffs' Human Action Methodology, such actions are classified as either knowledge or skill based. This human action methodology appropriately degrades the failure probability given that procedural guidance and/or training is limited or does not exist for a particular action it is, therefore, too conservative to restrict human actions when the degree of procedure guidance and training can be addressed by available methodologies. It is recommended that this statement be modified to indicate the need for the assessment of the degree of procedure guidance and training available.

4

ATTACllMENT (1)

COMMENTS ON DRAIT NUREG 1602, DRAIT REGULATORY GUIDES

~

DG 1961, DG 1062, DG 1965. AND DRAIT STANDARD REVIEW PLAN CIIAPTERS 3.9.7, 16.1, AND 19, ADDRESSING USE OF PROllAHILISTIC RISK ANALYSIS IN Tile REGULATORY PROCESS

2. In Section 2.1.5.1, page 2 24, of NUREG 1602 states: "At a minimum, plant walk throughs, interviews with plant personnel (e.g. training, maintenance, operators, shift supervisor, shift technical advisors), and procedure reviews are performed in identifying and selecting the human events for a PRA." It also adds that, " Observations of simulator exercises and maintenance crew performance can be used to provido additional information . . ." These stipulations, if meant to be a set of minimum requirements, are both impractical and unrealistic. With the number of human actions currently being used in PRAs (over 100 in Calvert Clifts' PRA), the additional time and

• resources required just to interview the recommended personnel is impractical. Interviews with knowledgeable opecators using well defined action descriptions and table top walk throughs are believed to be very effective. The above statement should be modined as such.

3. In Section 2.1.5.1, page 2 25, Screening / Excluding Iluman Events, states: "A screening analysis can be performed to identify and exclude these (non critical) events from detailed evaluation.

Ilowever, the screened human events should be reconsidered for every risk informed regulatory application of the PRA to ensure that all of the risk contributing actions are included in the spplication analysis." The intent of this statement is unclear. Is this statement focused on pre-initiator human errors? Is a documented assessment expected for each application?

4. In Section 2.1.5.1, page 2 25, Screening / Excluding Iluman Events, states: "lluman events, such as all pre initiators, generally cannot be excluded from consideration based on the argument that these events are included in the components hardware data. Many human events (such as miscalibration) occur rarely and are not necessarily reDected in the random failure data. Further, their effects can be subtle in that they impact multiple systems and, thus, can play a key factor in contributing to core damage." While we agree that pre-initiator human errors that impact multiple channels or t.ains should be addressed since they represent a potential common mode failure, it appears unreasonable to include this impact as a separate contributor to the failure rate of a single basis event. In addition, we believe that most, if not all, significant maintenance (including miscalibration) enors are captured in the hardware data. Since it is stated that many human error events are rare, this statement indicates that even if they are not captured in the failure rate data, they would likely be an insignificant contribution. Therefore, beyond the impact of multiple channels or systems, it does not appear that the level of effort to include these errors at a basic event level is warranted.

5. In Section 2.1.5.1, page 2 26 NUREG 1602 states: "The available time is accident specine and determined from engineering analyses which include actual time measurements derived from walk-throughs and simulator observation." This statement is apparently confusing the available time with the actual time. The available time should be based on engineering analyses or model structure that establishes the constraint for an action. The actual time is determined through the interview process that, at a minimum, should include a table top walk through.

6. In Section 2.1.5.1, page 2 27, NUREG 1602 states: "For post initiators, the human event should be evaluated relative to the specific context of the accident progression. Therefore, for different accident sequences, the human event is evaluated for each sequence." Since a human action has the possibility of being utilized in multiple sequences, possibly thousands, analyzing each particular use would be impractical. The recommendation should be clarified to say: "Therefore, a human failure probability should be bounding for all accident sequences for which it is used." The boundaries both define the action bein.g interviewed and establish *be boundary for its use in the plant model.

5 m

ATTACllMENT (1)

COMMENTS ON DRAIT NUREG 1602, DRAIT REGULATORY GUIDES DG.1061, DG.1062. DG 1965. AND DRAIT STANDARD REVIEW PLAN CilAPTERS 3.9.7 16.1, AND 19. ADDRESSING USE OF PROHAHILISTIC RISK ANALYSIS IN THE REGULATORY PROCESS 1 The bounding condition should also include the key electrical and/or mechanical supports required

.f by the operator to identify, diagnose, and perform the action.

I

7. In Section 2.1.5.1, page 2 27, NUREG 1602 states that errors made in performance by the original 4

operator can be " recovered" by the same operator and by other plant personnel (e.g., shin technical advisor). Total credit for all such " recoveries" should not exceed a factor of ten (higher credits should be identified andjustified). Baltimore Gas and Electric Company believes that this factor of ten seems too conservative for all such applications. It is suggested that, if a second redundant person (e.g., shin technical advisor) whose sole purpose is to do certain safety function status

checks on a set interval catches the initial error, then the factor of 10 can be exceeded. In fact, this second recovery action by the shin technical advisor should be able to be quantified as a separate indvidual actien as long as attention is given to the preceding related unsuccessful action. Other actions, where a second alarm may come in given that the initial indication went unnoticed, should be interviewed in such a manner so that the initial mistake has a consequence on the recovery action. The outcome of such an analysis may or may not be bounded by the factor of ten.

Accident Sequence Ouantification (Section 2.1.6)

In Section 2.1.6.1, page 2 30, Selecting Truncation Values states: "At least 95% of the total core damage frequency and 95% of the early and late release frequencies should be expressed in the model results. Also, it should be verified that lowering the truncation limit does not significantly increase the

model estimation of the total core damage and release frequencies." Truncation is required in order to make large fault trees and event trees solvable given the available computers and computer codes.

Baltimore Gas and Electric Company uses PLO, Inc.'s RISKMAN software. Cutoff frequencies speed the quantification of event trees, and make it possible to build very large trees, quantify the highest frequency sequences and bypass insignificant sequences. Cutoff frequencies can be set for each initiating event. RISKMAN calculates the frequency of sequences that fall below the cutoff frequency.

Tnis frequency can be a large percentage of the calculated CDF. Ilowever, the portion of these unaccounted sequences that lead to core damage is unknown. Therefore, stating that "95% of the CDF should be expressed" is a target that cannot be verified unless all sequences are quantified (at which point truncation would not be an issue). Stating that a truncation value that is four orders of magnitude lower than the final CDF is sunicient, is believed to be non-conservative. Using this method would significantly decrease Calved Cliffs Nuclear Power Plant's calculated CDF. Requiring truncation to be considered before and aner operator recovery is also inappropriate for the large event tree methodology.

Iluman actions are included as either top events within the event tree or basic event within a fault tree that is used to calculate the top event. Removing these actions would prove to be extremely difficult and time consuming. Ilowever, BGE does agree with the statement: "Also. it should be verified that lowering the truncation limit does not significantly increase the model estimation of total core damage and release frequencies." To ensure greater consistency, a numerical value should be assmiated with "significantly increased." Determination of this value may require some sensitivity analysis.

I Section 2.1.6.1, Estimating Uncertainties, page 2 31, addresses tue use of mean values and guidance on estimating uncertainties. It is BGE's opinion that the use of point estimates, with sensitivity analysis on key parameters, would be just as valuable and more workable. Performing uncertainty analysis for each application appears to be an unwarranted cost and will likely inhibit risk informed regulation.

6

ATIACilMENT f1)

COMMENTS ON DRAIT NUREG 1602, DRAIT REGULATORY GUIDES DG 106), DG 1962 DG 1965, AND DRAFT STANDARD REVIEW PLAN CilAPTERS 3.9.7, 16.1, AND 19, ADDRESSING USE OF PROHAlllLISTIC RISK ANALYSIS IN Tile REGULATORY PROCESS Section 2.1.6.4 Documentation, page 2 33, provides a list ofinformation regarding PRA quanti 0 cation that should be documented. This list of bullets goes well beyond the information needed to ensure the quality of the quantification process. In addition, the bullets appear to be focused towards the CAFTA

/conynner alded/mdi trce analysis] sonware code quantl0 cation process. The following changes are recommended:

Dullet 2: Remove "and accident classes."

Ilullet 4: Delete. ne documentation of human actions is addressed in Section 2.1.5.4. This bullet addresses the assessment of the results, which is better addressed when assessing the PRA for a particular application.

Ilullet 6: Delete. See Section 2.1.6.1 uncertainty comment.

Bullet 7: Delete. This bullet is a post pro:ess item its importance will vary in accordance with each application.

Bullet 8: This bullet appears to be sonware specific and not applicable to a RISKMAN quantification.

Dullet9: Maintaining a Ole of all sequences (currently greater than 100,000) appears reasonable.

Providing brief descriptions of these sequences is not. Therefore, delete the description requiremeat.

Bullet 11: Please denne "non recovery terms."

Dullet 12: Delete, nls bullet appears to be sonware specific.

PEEli REVIEW In NUREG 1602, Appendix H, PRA Peer Review, addresses the objectives and scope of a peer review.

This section implies that an extensive peer review should be performed with each application. This requirement is inconsistent with DG 1061's quality audit requirement and with the risk informed Technical Specification pilot. The quality audit requirement implies an audit on the PRA process not the application. For the Technical Speci0 cation pilot application, cross comparisons were performed in lieu of the detailed peer review outlined here.

It also appears that the line between independent review and peer review is unclear, ne peer review described is onerous and costly, and substantially overlaps the role of the independent reviewer, it is more appropriate to focus the peer review on the review of key assumptions, the methodology and the reasonableness of the conclusions.

Baltimore Gas and Electric Company recommends that NUREG-1602 be used as a tool for self-assessment performed by the licensee's PRA staff. We believe that high level peer reviews perfonned by PRA experts to assess the key assumptions, the methodology, and the reasonableness of the conclus'ons coupled with independent document reviews is a cost effective method to achieve a quality PRA. Peer reviewers who are experienced PRA practitioness, but who are not necessarily knowldgeable with the plant, are believed to be acceptable. The verincation of the model to the plant is the responsibility of the independent reviewers. We also believe that a peer review for each application is unnecessary and appears to exceed the 10 CFR Part 50, Appendix B requirements for safety-related work.

7

NITACllMENT (1)

COMMENTS ON DRAFT NUREG.1602, DRAFT REGULATORY GUIDES

'o

• DG 1061. DG.1062. DG 1965. AND DRAFT STANDARD REVIEW PLAN CilAl'TERS 3.9.7 16.1, AND 19, ADDRESSING USE OF PROllAlllLISTIC RISK ANAIASIS IN Tile REGULATORY PROCESS MAINTENANCE RULE Conspicuously absent from tiie framework of risk informed regulation are guidelines for the most utilized PRA application: on line maintenance risk assessment. In responss to a question at the workshop as to the quality of the PRA required for this analysis, it was stated that the absence of guidelines for on line maintenance risk assessment is a problem for the industry to resolve. Since on line maintenance is a necessity for cost effective operation, the design of a risk informed regulation without resolution of this issue appears to be a major weakness. The silence in this area muddles the regulatory requirement for having a PRA and contributes to the considerable plant to plant variation in the level of PRA resources. The development of a Configuration Risk Management Program for risk informed Technical Specifications appears to be an attempt to address this weakness. By placing this overhead only on risk informed Technical Specifications, it discourages the pursuit of these Technical Specifications since the cost of a quality Configuration Risk Management Program is high. A framework that does not address this issue will result in a division in the industry between PRA players and those that are not. Ilattimore Gas and Electric Company recommends that guidance on this issue be clearly stated.

~

8

--1