Text

Transcript of ACRS Ac/Dc Power Sys Reliability Subcommittee 900808 Meeting in Bethesda,Md.Pp 1-105.Viewgraphs Encl
	ML20056B022
Person / Time
Issue date:	08/08/1990
From:	; Advisory Committee on Reactor Safeguards
To:	;
References
	ACRS-T-1807, NUDOCS 9008130234
	Download: ML20056B022 (133)
	v • d • e

{{#Wiki_filter:j ?. IW ~~ RCRSW/967! CRIGLNAL 1 9,Jr OFFICIAL TRANSCRIPT OF PROCEEDINGS a a g .n ,to I i 9 Jl 5 t ,P f 6 r.. i 'l 1 ,w 'i r 1 U.S. Nuclear Regulatory Commission i U Advisory Committee on Reactor Safeguards . Tkle.. AC/DC Power Systems Reliability. Subcommittee 4 i'. q Docket No.' l-e ' h' i m.dONi Bethesda, Maryland o.. i \\ chymf Wednesday, August 8, 1990 PAGES: 1 - 105 .i.

(

f b ,?- of)

L-ACR8 0FFICE COPY 1

E Jo No:. Remove from AC38 0:" ice 3 H. l 6;l L: ANN RILEY & ASSOCIKfES,13D. h 1612 K St. N.W, Suite 300 washington, D.C. 20006 (202) 293-3950 ooos; 30 :u w,s o.3 [U h 5 [ki, R E T-1807 FDR l-M '!)]; il

l'B lif x

O

x-3

-l 4 - PUBLIC NOTICE BY THE .5 UNITED-STATES NUCLEAR REGULATORY COMMISSION'S i .h ' ADVISORY COMMITTEE ON' REACTOR SAFEGUARDS .i 6 .n 7-l [8 DATE: Wednesday August 8, 1990 4 - i g-9 101 > i j 11 1 112 ,a

)

,y '13 The contents of this transcript of the 14 proceedings'of the United States. Nuclear Regulatory -j 4 X: 15 Commission's. Advisory Committee on: Reactor. Safeguards, . ednesday, August-8. 1990 16. (date)' W .17 : as reported herein,.are a record of:the' discussions recorded at .i l 18: the meeting. held on the above date. . \\ ^: 19 -This transcript has not been. reviewed, corrected ( 20 or' edited,-and:it may contain inaccuracies. 4 u-

22 J'

23 24 -( '25 g 3-q'. j;..

._.----a---------x--.-

A, p-i .3, -1 i l ll j UNITED STATES OF' AMERICA' L a 2' , ni 3 NUCLEAR REGULATORY ' COMMISSION .a. 4 Sj ~ ADVISORY COMMITTEE: ON REACTOR SAFEGUARDS in i G: -\\ - AC/DC Power Systems Reliability Subcommittee 8-i v 3 9-Nuclear Regulatory Commission j 10 7920 Norfolk Avenue 11 Bethesda, Maryland p -- 12 1 13; Wednesday, August 8, 1990. 14' 15 The-above-entitled proceedings commenced at 8:30 f. --16 o' clock a.m., pursuant to notice, - Charles Wylie, j 17 Subcommittee Chairman, presiding. 18- -1 19 -PRESENT FOR THE SUBCOMMITTEE: 20 C. Wylie c21-W. Kerr 22 H. Lewis 23 C. Michelson 24-J. Carroll ( 25

2 l-l 1 APPEARANCES (continued): "O 3 AIEO PRESENT: 4 5 M. El Zeftawy, Cognizant ACRS Staff Member 6 7 8 9 10 11 12 I'l 13 14 15 16 17 18 k 19 20-21 22 23 ( O 25

i 3 1~ PROCEEDINGS } t' 'i, t \\) 2 (8:30 a.m.] \\ 3 MR. WYLIE: Suppose we get ready. We've got 4 another meeting this afternoon following this one, so we 5 havefto move right along on this meeting here. 6 The meeting will come to order. This is a meeting 7 of the Advisory Committee on Reactor Safeguards, 8 Subcommittee on AC/DC Power Systems Reliability. I am 9 Charlie Wylie, Chairman of the Subcommittee. 10 The ACRS Members in attendance today are Messrs. 11 Harold Lewis, James Carroll, and Dr. Bill Kerr. Carlyle it Michelcon will join us shortly. ^r U 13 The purpose of the meeting is to review the NRC 14 and NUMARC Proposed Resolution of Generic Issue B-56, Diesel 15 Generator Reliability. Dr. E)-Zeftawy is the Cognizant ACRS 16 Staff Member for this meeting. 17 The rules for participation in today's meeting 18 have been announced as part of the Notice of this meeting 19 previously published in the Federal Register on July 27, 20 1990. 21 A transcript of the meeting is being kept and will 22 be made available as stated in the Federal Register Natice. 23 It is requested that each speaker first identify himself or 24 hercelf and speak with sufficient clarity and volume so that 25 he or she can be readily heard.

.Vy,e 4 1 We have received no written comments or requestu to make p; J ) .k/ 2-oral statraents from members of the public. I think the i' 3 status report provided the Subcommittee by Dr. El-Zeftawy is 4 a very good summary of the past activities and situation on S the diesel generator reliability. 6 If the Subcommittee will recall, we met with the 7 staff and industry, NUMARC, and reviewed the activities on 8 the Proposed Reg Guide 339, and NUMARC guidelines for 9 addressing station blackout, including Appendix D, in 10 October, October 6, and then we had a briefing of the Full 11 Committee and -- correction -- that was October 2nd. Then 12 we briefed the Full Committee on October 6th, and then again ( ) 13 on February 8, the staff and NUMARC. 14 At that time, things looked favorable. There were 15 some differences and the Committee felt that the staff would I 16 be able to work those trerences out. So I said we'd have 17 a complete Reg Guide endorsing the NUMARC guideline, and so 18 urged the staff and NUMARC to work together toward that end. 19 I understand the staff feels that due to 20 activities that have progressed since that meeting and to 21 revisions, that that is not the staff's present feeling. 22 Th3y can endorse the present version that it must contain l' '23 additional guidance and on-site diesel generator reporting .f^} 24 requirements. g 25 Tne staff and NUMARC are here to address this

D

5 1 issue, and possibly they could identify the differences in w-) 2 why the staff feels, at this point, they cannot endorse the 3 NUMARC guidance. 4 The Subcommittee can consider what action we 5 should recommend to the Full Committee and what should be 6 brought to the Full Committee tomorrow afternoon at 1:30 to 7 3:15 on this issue. So I would ask the Subcommitte& 8 members, as we progress this morning, to make note of what 9 they think should be brought before the Full Jommittee and 10 what actions we should recommend that the Full Committee 11 take. i 12 With that, Warren, I guess we'll turn it over,to ( ) 13 you. 14 MR. MINNERS: Thank you, i 15 MR. WYLIE: Go ahead, Warren. 16 MR. MINNERS: We're here-to talk about diesel r 17 generator reliability programs. The chronology is a long 18 one. Le really started out before the Station Blackout Rule 19 was issued, we really started to get going on this thing, 20 and we've met with you and CRGR and NUMARC and had finally 21 thought we had come to an agreement last winter. 3 22 We were then going to just endorse the NUMARC 23 guidance. So we were waiting for NUMARC to submit their ~ 24 guidance, which they did in May. At that time, we found 25 that they had made significant changes from what we thought

6 ~ 1 we had agreed upon back in November and December. I will lNf 2 explain later'what the changes were. 3 We took that to CRGR recently. As you have in 4 front of you, we have some recommendations from CRGR that 5 we're going to adopt. I will explain what they are. So 6 today and tomorrow, we're meeting with the ACRS to explain 7 those positions. 8 (Slide.) 9 MR. MINNERS: A quick overview of the Reg Guide. 10 It takes the old Reg Guide 1.108, which was really a lot of 'll design guidance, and incorporates it together with a 12 reliability program, which is the new stuff, into one ( ) 13 document. 14 The old stuff is not boing backfit. The 1.108 15 stuff is implemented by some' plants and not by others, but 16 we're not recommending that that be backfit because it's 17 basically design kind of stuff and it's really impractical. 18 But we are recommending and requesting that the 19 reliability program, the new stuff in the Guide, be backfit 20 to all plants. 21 MR. KERR: When the staff talks about a 22 reliability program, this means a program, something or 23 other, that is designed to achieve reliability? I'm curious 24 as to what one means by that phrase " reliability program." 25 MR. MINNERS: As we mean it, it's an overall

) 1 program with seven elements which we think are necessary to I) 2 achieve the desired target reliabilities and to give j i 3 adequate assurance that that reliability has been attained. 4 MR. KERRt So it isn't enough to simply 5 demonstrate that one has reliability and one also has to 6 have a program in place which is aimed to achieve that 7 reliability. 8 MR. MINNERS: Correct. 9 MR. CARROLL Warren, I recall from our past s 10 discussions that there is some controversy about whether 11 this should include design issues. Did I understand you 12 corractly that design issues are separate from what we're I 13 ta.'. king about today? 14 MR. MINNERS: That's right. This Reg Guide will 15 apply to future plants. So the design portion of the Reg 16-Guide will apply to future plants. 17 MR. CARROLL: So there is a design portion of the 18 Reg Guide? 19 MR. MINNERS: Yes. It's the old Reg Guide 1.108, 20 which I think is basically what I would characterize as 21 design, although there is more than that in it. That's a 22 rough characterization. But we don't want to backfit that l 23 to plants, because that doesn't make a lot of sense. 24 MR. KEhR: In wnat sense is the term "backfit" 25 - being used here, because I think a backfit in terms of a l l w ---ww u- .m-,- ,~+

I 8 1 regulation, but a Reg Guide is not a regulation. So what 7-G i 2 does "backfit" nean? l 3 MR. MINNERSt The backfit was done through the 4 Station Blackout Rule. That's where the backfit came in. 5 That's our authority for backfitting this. The Backfit Rule 6 anticipated having a reliability program so that you could 7 assure that the target reliabilities of the Blackout Rule 8 were attained. 9 MR. KERRt What do you mean by " anticipated" that? i 10 MR. MINNERSt I don't know. Just that. That 11 that's the basis of the Blackout Rule, is that there would 12 be a reliability program to assure that you had the target 13 reliabilities. 14 MR. KERRt But it didn't have to be a Reg Guide. i 15 It could be some other program that -- I'm trying to 16 understand what is meant by backfit where it is not applied 17 to a regulation, i 18 MR. MINNERS: I'm trying to tell you that it does 19 apply to regulation. The Station Blackout Rule was backfit 20 to all plants. 21 MR. KERRt But you said that the Reg Guide, some 22 point of the Reg Guide would not be backfit. I'm trying to 23 understand what it means to backfit a Reg Guide. . T 24 MR. MINNERS: The Station Blackout Rule only b(g 25 talked about reliability programs, and that's the part we're

t ) O d .) 1 backfitting. We thought we were going to simplify the $'~)' 2 process by putting all the diesel generator requirements q ~' l 3 into one Reg Guide, but some of those were not backfit by 4 the Blackout Rule. 5 MR. KERR But a Reg Guide can't be a requirement 1 6 since it's not a rule. 7 MR. MINNERS: I didn't say -- it's not a 8 requirement. The Reg Guide is as it always ist it's 9 guidance to the industry of an acceptable method of 10 complying with the rule. 11 MR. KERR: Unless I'm mistaken, I thought you said

g 12 you were going to put all the requirements for diesel L h,7 13 generators in the Reg Guide.

11 14 MR. MINNERS: Put all of the guidance for diesel 15 generators in one Reg Guide. I misspoke. Right. I used 16 requirements too loosely. 17 MR. KERR: And I'm back to my question about what r. 18 does it mean to backfit a Reg Guide. 19 MR. MINNERS: It means just that; that if the -- 20 we're not backfitting the Reg Guide. We're backfitting the w 21 rule, and this Reg Guide is inherpreting how the staff would g 22 find an acceptable 'mplementation of that rule. 23 MR. KERR: But not the only acceptable one, MR. MINNERS: Correct. People can come in and { } 24 L 25 present something. They also will talk about. But the

i o 10 1-purpose of having a -- ,s i A /- '2 MR. KERR Wait a minute. You said you talk 3 about. Surely you would do more than talk about it if there 4 are alternatives. 5 MR. MINNERS: You mean that we will accept it? 6 MR. KERR Not but that -- you give me an i 7. impression that you wouldn't accept anything other than the ] 8 Reg Guide. 9 MR. MINNERS: That's not what I said. I said that 1 10 we would talk about and then we would decide whether to i 11 accept it, modify it, or reject it. ] 12 MR. KERR Okay. l t( ) 13 MR. MINNERSt I think the first step would be to 14 talk about it. But the whole purpose of sitting down with 15 NUMARC over the past 26 months that we've been doing this 16 thing is to try to get a set of guidance that the industry 17 would agree to and we wouldn't have to have a lot of rsview 18 by the staff, with people coming in with their own 19 particular interpretation. l 20 That's why we have worked so hard with industry to 21 try to get something that would be acceptable to both of us, 3 i 22 and we could save time for both of us. That's the iden I 23 here. 24 So in that vein, Item 5 here is why what we put (r~)N 25 into the Guide incorporates proven industry practices and is ( l l l r r

11 i consistent with the NUMARC Appendix D and their related g e ~2 Topical Report. -3 So what we're doing, we think we have agreement 4 with industry on, or had agreement with industry on. 5 MR. KERR What does " proven industry practice" 6 mean? l i 7 MR. MINNERS: It's practices that the people who 8 have good diesel generator performance use to maintain that 9 performatice. 10 MR. KERR Okay. 11 MR. MINNERS: It's not a mathematical proof, no. 12 It's the general tern; that it's been used in practice and 13 people who are experts in diesel generators say, hey, that's 14 a good thing to do. And it's a matter of opinion. 15 (Slide.) 16 MR. MINNERS: This is where the differences come i 17 up. This is a comparison of the Reg Guide versus the May 18 edition and the official edition of the NUMARC-8700 Appendix 19 D. All of this stuff up here is the stuff that we were 20 talking about, which ic the design stuff which is not being 21' backfit. l: 22 That's why Appendix D doesn't contain that; 23 because NUMARC said they were only going to do reliability, 24 which is correct. We all have agreement here. This part is 25 the same, because this is definitions. We put all the 1 L

12 1 definitions in one place. Some of the definitions apply the f 2 design stuff and some apply to the operational reliability 3 stuff. 4 We have good agreement here on the goals and i 5. monitoring and our recordkeeping. Now, where we have fallen 6-out is on reporting criteria and on the details of the { 7 reliability program. 8 We have, in the Guide, seven elements of the o 9 diesel generator reliability program. The industry has 10' chosen to only put one of those elements into Appendix D. 11 The remaining elements and explanation of those elements 12 they have put into a Topical Report, which they have not -13 submitted and don't want to, as I understand, submit to the 14 Commission, but it will be available far use by utilities. 15 So that's our basic difference of whether this 16 information should be in a Reg Cuide or whether it should 17 not be officially endorsed.., ' 18 MR. KERR: Now, if it is put in a Reg Guide and a 19 licensee says I accept that Reg Guide, it then has the 20 defect or relationship or rule against which the licensee is-21 inspected. 22 MR. MINNERS: If he commits to it, yes. I don't t 23 know. I'd have to get my lawyer down here to tell you what .p-24 the enforceability of a Reg Guide is or what the 25 enforceability of commitments are. But I think the staff L w, \\ i

.i i i 13 I would generally say, yes, they committ9d to it, you ought to N 2 follow it. j 3' MR. KERR: A licensee could then be cited or fined 4 for not idhering to -- 5 MR. MINNERS: You'll have to get the lawyers down 6 here to answer those kinds of questions, Dr. Kerr. I don't .7 know. 1 8 MR. KERR: It seems to me that you need to know if 9-you're going to understand why the licensees may object or 10 not object to it, because it is something with which they 11 are going to have to deal. 12 MR. MINNERS: Well, fine, but I don't think the I( ) 13 legal question is the one. The question is what are they 14 going to -- 15 MR. KERR: I'm not asking a legal question. I'm 16 asking a question of practice. What happens, what is the a 17 effect of this rule on the licensee. 18 MR. MINNERS: That's correct, and that's what we 19 have tried to consider and we've tried to say, hey, what we 20 vant to do is to have guidance which you guys agree is tae 21 proper thing to do; that together we agree this is the thing 22 for the industry to do in operating its diesels. l 23 MR. KERR: So in addition to reliable diesels, you 24 want a specific method of doing something which you think is ( 25 necessary to achieve that. It isn't enough to just have the ~. ,,... - ~

14 1 reliable diesels. 2 MR. MINNERS: I don't understand your question. 3 MR. KERR: It seems to me that you are asking for 4 more than diesel reliability. You are asking for a specific 5 approach which, in the view of the staff, should achieve 6 that reliability. 7 MR. MINNERS: You have to have assurance that you 8 have that reliability. There is no way of knowing if you 9 have any reliability, because you don't know. There are 10 only ways of monitoring and testing the diesels which gives 1 you indications of what the truth is. 1 12 You don't know what the diesel generator l( ) 13 reliability is and you will never know. You can only have 14 programs and testing which will give you some degree of 15 assurance that the reliability is there. 16 MR. KERR: I'll wait and see what happens. I'm 17 just trying to understand the difference in viewpoint 18 between -- 19 MR. MINNERS: Your point is exactly the point I'm 20 trying to make. Let me go through my story and I think I 21 will answer your question, because tha'e is the question to 22 ask. 23 MR. KERR: Thank you. 24 MR. CARROLL: Just one clarification on that one. g R25 6.1 through 6.7 formerly were in the NUMARC document, is 1 1

l 15 l 1 that correct? . Q, N-2 MR. MINNERS: Formerly were in the NUMARC 3 document. 4 MR. WYLIE: But now only 6.1. 5 MR. MINNERS: Now only 6.1 is in the Appendix D 6 and the other ones are in this Topical Report, which is not 7 here. 8 (Slide.) 9 MR. MINNERS: This just kind of expands on the j 10 elements of the program. We, as I say, agree on Item 1, L 11 which is the monitoring program. We have worked that out in L 12 detail, and it's the other ones. NUMARC has items 2 through ~ k ) 13 7 listed in the introduction to their Appendix D, but they j '14 don't have any explanation or examples of how to implement 15 those other six elements. 16 So I think we agree on the seven elements. It's l 4 L 17 the detailed information that we may have some disagreement 1 18 on. 19 MR. KERR: Again, I apologize, but the monitoring 20 of the reliability -- l 21 MR. MINNERS: Right. ( 22 MR. KERR: -- would give some evidence that a 23 reliability is being achieved, would it not? /~'\\ 24 MR. MINNERS: Yes. i1& 25 MR. KERR: Okay. l l L 1

16 1 MR. MINNERS: Would give some evidence. I want to { 'i s-2 go through that. 3 (Slide.) 4 MR. MINNERS: This is the part that we and the S industry agree on, which is the monitoring part, and this is l 6 just a logic diagram of this. Depending if you have no l 7 failures, you don't ignore them; you still look at root i b 8 causes and corrective actions. 9 If you exceed one of the trigger values, which I L 10 will get to later, you do a little more. You go back and 11 look at maintenance history and other failures and failure i 12 patterns. And then if you exceed two of the trigger values, r h 13 you even go more into it. 14 So none of this involves the NRC. The thrust of 15 this-program is to try to get us out of diesel generator -16 reliability. We're turning it over to industry. We're 17 trying to set up a program that we both agree upon. And if 18 people do this, we shouldn't have to get involved at all. h 19 There is g ing to be no -- you don't have to 20 report to the NRC in any of this stuff. I guess the 21 Resident Inspector can go in and inspect what he wants to 22 see if you're doing it, but we don't have any inspection 23 program to go out and look at this stuff. /"'g. 24 We do have a reporting requirement that if you get V 25 into what we call a problem diesel, which we think is a case

l 1 17-i .i i 1 in which the program has degraded very, very badly, then you Oi 1 js,/ 2 do have to report to the NRC. 4 4 3 But if you're in this normal thing -- and you can 4 go quite far down here. You can hit two of these triggers, i 5 which means you can have some fairly bad reliability, and 6 you still don't have to get in contact with us. 7 So we're trying to stay out of the game. We're 8 trying to turn it over to the industry and let them do their 9 thing. That's why we're working so hard to try to get a 10 program that we can both agree upon. This is going to be L 11 something that's going to be set in place and we're not 12 going to look at it again. I( ) 13 We're not going to monitor diesel generator 14 reliability to see how these people are doing. So we think 15 it's fairly important to be sure we've got it right. 16 (Slide.] 17: MR. MINNERS: There's a problem with monitoring. 18' The concern that the utilities have, and correctly so, is 19 with the false alarm rate. This says that if you really. 20 truly had a steady-state reliability of 98 percent on your ' 21 ' diesel system -- and we're talking about systems here, we're .22 not talking about individual diesels.. 23 As you put more and more demands on the diesel, gr 3 24 there is some finite probability that you'll trip the alarm V 25 rate. This is the ono that we are proposing in the Reg

l 18 j 1 Guide, which is three out of 20 failures, or five out of 50, i gf's-2 or eight out of 100. 3 That would give you the mild alerts the middle box i 4 of the previous diagram. We also have two triggers. If you 5 have five out of 50 and eight out of 100, that would put you 6 in the third box. That's a strong alert and you have to do 7 more. That's what these lines are in here. 8 So for the one trigger value, there is a finite 9 probability that you will pick that up. That's what they're 10 worried about. That wts for 98 percent reliability. 11 MR. KERR: Warren, I'm sort of slow on this. 12 Would you explain to me what it is that they're worried l O bg. 13 about, again? 14 KR. MINNERS: They're worried that when they have 15 98 percent reliability, which is okay -- they're only 16 shooting for 95 -- if they have 98 percent reliability, 17 there's a-finite chance they're going to hit one of the 18 triggers and have to go through one of their alert cycles, 7 19 and they don't want to do that too many times. They don't 20 want to have too many false alarms. That's what this is 21 worried about. 22 So if you have a very high reliability, I think 23 this is a reasonable probability that you're going to have a 24 false alarm. They don't want to have too many false alarms [ 25 and have to go through the extra steps in a mild alert or in

I 19 1 a severe alert. 2 They shouldn't have to do that. You don't want to J 1 3 have falso alarms in your monitoring program. J 4 MR. KERR: Is a falso alarm the same thing as a 5 statistical fluctuation?

u 6

MR. MINNERS: Yes. This is due to the randomness 7 of things. This is just due to the randomness of the thing. 8 MR. KERR: That's what I was trying to find out. e 9 Thank you. 10 (Slide.] 11 MR. MINNERS: Now, things get worse, of course, if 12-the diesel generator system is really running at 95 percent, I( ) 13 if that's the true reliability, the randomness here -- I 14 mean, now you get some pretty high probabilities that you're 15 going to hit one of the alert levels, even though you're at 16 the target reliability. 17 So in choosing the test frequency and in choosing 18 these alert levels, we try to balance them out, because if 19 you have people test too frequently,-you're going to get l 20 more of these things. 21 So we had many long discussions of what numbers to 22 pick.out. This isn't strictly mathematical and we made what 23 we thought was a proper balance. So these are the numbers g L 24 that we picked, but it's going to give you some probability y 25 of false alarms. l -. l L

I[ w l 20 lu MR. LEWIS: Warren, I wonder if I could ask a 7s ) > (( 2 question. You've used the word " mathematical" twice as if i 1 3 it were a dirty word. I just want to put on the record I i 4 don't think it's a dirty word. 5 MR. MINNERS: I didn't mean to say it's a dirty i 6-word. i 7 MR. LEWIS: I know you didn't. l 8 MR. MINNERS: I'm just trying to say we didn't do 1 9 it in a strict rigorous manner. 10 MR. LEWIS: I know, but you seem proud of not i 11 having done it that way. That's a minor point. The major 12 point is that the failures of the diesel ;; aerators seem to t l E, ) 13 vary very roughly follow exponential model. That's a i 14 mathematical comment of which I'm not asi,emed. IS' Given that, there are expert,on the theory of 16 reliability who will tell you what the optimum testing ( 17. ' program is to produce a given level of reliability with an 18 exponential model. Give them a model, they'll give you a c 19 method. 20 Were such people brought into this business? 21-MR. MINNERS: Yes, sir. [ 22 MR. LEWIS: From within or outside NRC? 23-MR. MINNERS: We had our contractors work with the 24 EPRI contractors. \\ 25. MR. LEWIS: That doesn't help me. Did they, in e

21 1 fact, use people who were experts on the -- 7x k,]' 2 MR. MINNERS: Yes. 3 MR. LEWIS: -- theory of reliability? 4 MR. SERKIZ We had one fellow from Colorado. He 5 had about 40 years experience with diesels. We had another 6 fellow with Lake Engineering with about 30 years experience, j 7 MR. LEWIS Those are institutions. I just i 8 wondered whether individuals know the business. ) 9 MR. WYLIE That doesn't answer the question. 10 MR. CARROLL: I don't think you understood his 11 7uestion. These are diesel experts as opposed to L 12 reliability experts? b(_)x 13 MR. MINNERS: We also had statisticians working on L L 14 this, yes. l 15 MR. SERKIZ: Al Serkit from Research. In the 16 course of the 20-some-odd months that we interacted with 17 NUMARC, NUMARC had statistical analysis people and people 18 that were knowledgeable in the operations of diesels, and 19 also were knowledgeable in what the patterns were. 20 I believe it was Aaron Engineering that was the 21 principal contractor, plus EPRI people worked with them. p 22 Basically, what they did was develop a calculation or a mathematical approach that was based on Monte Carlo analysis 23 l-p } 24 techniques, and I won't try to explain those to you because 25 I'm not a mathematician in that degree. l

22 1 We utilized people from SAIC that had done work 2 for Brookhaven and continue to do work in the areas of f 3 reliability monitoring and this sort of thing. e 4 Now, both parties went off on their independent 5 approaches, having their own mathematical similarities, and 6 we had several meetings, at least three meetings in where 7 the comparisons were made by two different calculational 8 methods. 9 When finally all was said and done, what we 10 developed was a set of trigger values which Warren showed 11 you in -- f 12 MR. MINNERS: They're right up here, A1. These 13 are the trigger values. 14 MR. SERKIZ: These were the calculations that led 15 up to the selection of the trigger values that are shown in 16 your Slide 6, which the best way that I can express 'it was u 17 an approach that says, okay, if you are striving for a 18 target reliability on a nuclear unit of 95 percent, you take i 19 a look at the failure counts or the trigger values of the L' 20 three out of 20 or five out of 50 and eight out of 100. 21 There is a lot of documentation that went into l 22 this. So what we did, b,,th the NUMARC people utilized EPRI 23 and Aaron Engineering and we utilized SAIC. The 24 mathematical and/or statistical analyses were done by these 25 people. We came back and we finally agreed that the best, nl, i li l

f 23 1 the most simplistic approach that va could lay down are -s 2 these trigger values that you see here to develop a course 3 of action for taking actions, which is in Figure 1. 4 Figure 1 appears also in NUMARC's Appendix D, May 5 2, 1990. So the mathematical people were there. The .6 statistics people were there. We can provide the Committee 7 with names, companies or whatever. 8 MR. LEWIS: Well, companies don't do me any good. 9 You've overwhelmed me with companies. 10 MR. SERKIZ: Ernie Lofgren, John Boccios, people I 11 like that wave involved from Brookhaven, as well as from 12 SAIC. I can't grab them all off the top of my head, but I <~ ( ) 13 can look them up and provide them to-you, sir. 14 MR. LEWIS: It would be nice to know the 15 credentials -- + 16 MR. SERKIZ: I certainly will. 17 MR. LEWIS: -- because you've said at one point 1 18 that these trigger levels were chosen as the most effective 19 and most simplistic, I think, was the word you used. 20 'MR. SERKIZ: That's correct, sir. MR. LEWIS: And those may not be the same thing, 22 of course. In fact, they're simplistic, I agree instantly. 23 Three numbers I can remember. But whether that is the best ) 24 .way to assure a statistical level of probability, I rather l 25' doubt with an exponential model, but it would be nice to I m m e ,e

24 1 know that somebody who knows the theory of reliability, who 2 is not a diesel expert, got involved somehow. That's all I 3 really needed. 4 MR. SERKIZ: Well, I would suggest that the 5' Committee, there's a NUREG CR-4810 which is titled Issues 6 and Approaches for Using Reliability Alort Systems to Detect 7 Component Reliability Degradation, which was prepared by 8 SAIC and will be published shortly. It's this report and it 9 contains a lot of that type of information. 10 I can leave this with you if you want to make a 11 copy. 12 MR. LEWIS: It would be nice to look at it, . pl} 13 j because -- I 14 MR. SERKIZ: This is one approach. The minutes of 15 the meetings that transpired, the case that Warren noted on 16 Slide 2, are also available, and there are a lot of examples 17 that go with those and the parties that were involved. 18 MR. LEWIS: I don't need to be overwhelmed with l 19 paper. All I need to know is that somebody who knows 20 something about the theory of reliability,.which is a l '21 mathematical theory, got involved in this game. That's all l 22 I need to know. 23 MR. MINNERS: On both sides. - (f-w 24 MR. SERKIZ: Both parties 25 MR. MINNERS: Both the NUMARC people and the ---e_a,, w-4 y ,e

l 25 i ) 1 staff, through their subcontractors. 2 MR. LEWIS: I didn't know staff had people in that 3 category. 4 MR. MINNERS: Staff subcontractors, sir. 5 MR. LEWIS: That's different. 6 MR. SERKIZ: We do have a statistician. We have 7 Lee Aberton. 8 MR. LEWIS: I know. 9 (Slide.) 10-MR. MINNERS: Now, the other side of that coin is 11 can you -- if you have a drop in reliability, can you detect 12 it, depending on your alert levels. This is a graph that ( ) 13 shows that, and this is the probability of detecting a drop 14 using these alert levels, a drop in reliability from 98 to 15 92 percent. 16 You can see that when you have small numbers of 17 demands, there is a pretty good probability that you won't 18 detect this drop in' reliability. Obviously, when you get 19 way out here, yes, sure, you're going to pick it up. 20 But on monthly testing and if you have'a two-21 diesel plant, 25 tests will take you a year. So with a year 22 of testing, you still have something like a 50 percent 23 probability of not detecting, y 24 MR. CARROLL: So a demand is either a test or a 25 real start signal.

1 26 i 1 MR..MINNERS: Right. And they are defined in the 2 Reg Guide, what a demand is. 3 .MR. LEWIS: Warren, this is the sort of thing that 4 troubles me a little bit, because when you speak of 5 detecting a drop in reliability where the overall 6 reliability level is something like 97 percent or somethir.g 7 like that, which means on the average of a failure every 30 8 tries, on the average, you sort of cannot, in principal, 9 detect a drop from 97-and-a-half to 95 by looking at 25 or 10 30 demands. You just can't do it. 11 MR. MINNERS: Correct. 12 MR. LEWIS: When you even speak of detection, you 13 are violating certain statistical theorems, because when you 14 say " detect," you're assigning a probability that something 15 has, l'n fact, gone down or whether you've seen a statistical 16 fluctuation. l 17 -{ These are not trivial statistical questions. 18 They're things that people know something about. I don't 19 aven understand the term " detection response." 20 MR. MINNERS: Detection response just says that 21 you would actually reach these alert levels; three out of 20 22 or five out of 50 or eight out of 100 in 25 tests. In 25 23 demands, there are 50-60 percent probability that you O 24 wouldn't get one of these. V 25 MR. LEWIS: You can certainly calculate the .~ ~

l. .i. [ r 27 1 1 probability that you reach a given trigger level in a ( ) \\~/ 2 certain number of demands as a function of the -- 3 MR. MINNERS: 7ad that's all this is, i 4 MR. LEWIS: -- underlying reliability level. What 5 .you cannot do is go the other way. So there are deep 6 questions here. I don't think this is the right place to 7. talk about them,. but I think it is wrong to pretend that 8 they're not deep questions. 9 MR. MINNERS: Because of this probability of not 10 detection of response, we believe you need what we call a 11 reliability program, and those are the seven elements that f 12 we showed before, to fill this gap. Q hq,j l's If I want to be able to go out to the public and 14 say, yes, I'm meeting the target reliability with reasonable 15 . assurance, whatever that means, we think we need no't only a 36 monitoring program that has these alert levels in it, but we 17 need a reliability program that has those seven elements in 18 it to be assured that when we have failures that are less e 19 than the alert levels, that something is done about those, t 20 and even when you have no failures, that there's a proper. 21 maintenance and people are watching what they're doing. 22 That's all a reliability program is supposed to 23 do, is fill in this gap to try to give you assurance, give y;['T 24 you more assurance that you have your target reliabilities 25 with these testing frequencies and these alert levels. ,.w.; ,..,y...

. -.= i 28 l 1 MR. KERR: There is, somewhere, some demonstration /~N i \\- 2 that that assurance program will produce the results that -1 3 you want to achieve. 4 MR. MINNERS: No. Do you mean you want -- what do l 5 you mean by " assurance?" 1 6 MR. KERR: You used the term " assurance," Warren. 7 I was just using the same word you used. i 8 MR. MINNERS: You used the word " demonstration." i 9 What do you mean by " demonstration?" 10 MR. KERR: I say that there is a demonstration 11 that this assurance program will achieve -- you have some 12 reason to think, some data, something or other, that leads ~h ) 13 you to believe that this program of maintenance or whatever 14 will make it more likely -- ( I 11 5 MR. MINNERS: Yes. o 16 MR. KERR: -- than just -- 17 MR. MINNERS: Yes. -( 18 MR. KERR: -- tests. 19 MR. MINNERS: And I have that data. 20 MR. KERR: Okay. 21 MR. MINNERS: But this is more based on the y 1 22 opinion and judgment of diesel generator experts who say, 23 hey, this is what you should do to assure diesel generator [ 24 reliability. 25 (Slide.) \\; I

~ l 29 1 MR. MINNERS: We don't think that diesel generator 2 reliability is bad. In fact, we think diesel generator 3 reliability has been improving and is quite good. The 4 industry has reported that the diesel generator reliability ] 5 ~ is 97-98 percent, the diesel generator system. 6 We haven't independently verified this, but I 7 don't have any reason to disbelieve that. But when you do 8 go look at the detailed data, you do find that there is 9 always a small number of plants that fall below a L 10 reliability of, say, 95 percent, which is the lowest target 11 reliability in the rule. L l 12 There are lots of operating history which shows ( ) 13 that there are problems in diesels,. and some of the plants - 14 - Vogtle is an example, Cooper, Cook -- 15 MR. KERR: What was it at Vogtle that demonstrated ^ 16 that they were belCW 95 percent? l 17 MR. MINNERS: The diesel didn't start. l 18 MR. KERR But a 95 percent reliability doesn't li L 19 say that a diesel starts every time, does it? Is that the l [ 20 answer, that it's being used to demonstrate -- L 21 MR. MINNERS: That is an incident which is being j 22 us9d to support the contention -- I 23 MR. KERR: At Vogtle, that's the one that you're l-24 referring to. jg sg .Q) 25 MR. MINNERS: That's the particular one we're l..

-;( 30 (:- ^ l' looking at,fyes._ ,2 -MR. KERR: Okay. 3 -MR. MINNERS: But there is a long history of that. 1-4' 'MR. CARROLL: Haven't they had a long history of 5 problems of -- b C, MR. MINNERS: They've had a long history of 7 maintenance problems with their diesels. i 8 MR. CARROLL: -- pressure temperature switches? 9 MR. MINNERS: Yes.- 10 MR. KERR: Fn it's more than that one' incident.. 11 MR..MINNERS: Well, that was the culmination of-12 the long' history. We're going to be using the IIT Report- '13. which says here was the problem and here's the long-history -14 behind the-problem. 15 MR.-KERR: And Cooper had a similar long history?- 16 MR. MINNERS: Let me get those.. I've got a slide- ~ l17 on those, y '18 (Slide.) 19 MR. MINNERS: Here is the average diesel generator-20 reliability. This is an industry average for 1985, 1986, 1 21 1987, and 1988, and it's well within 95 percent. It's up 22' around 98 percent. No problem with that at all. g 23 This is the distribution over the 1986 to 1988 24 period industry-wide of diesel generator.reliabilities. 25 over this four-year period or three-year period, I guess, _ - - ~ - - - _ - _ _ _. -. _ _ _ _. ~

1 ',o 31 1 yestLeverybody was above 95 percent with a_three-year 4 L 2, average, according to this data, which, I think, agrees-3 that, in general, diesel generator reliability is good. 4 MR. KERR: Does that mean that the stations to I 5 which you referred; Vogtle, Cooper, cook, Calvert Cliffs, ~6 -and Zion were within this group or what you're alluding to 7 happened after 1988? 8 MR. MINNERS: Yes. That's an industry average, so 9 I presume that was in the data. 10 MR. KERR: I thought what you showed me earlier 11 was a number of-plants, like 14, that had an unreliability 12 and 11'had another one, and all of the plants that industry - k ) 13 . reported-on were within the 95 percent. 14 MR. MINNERS: Over the three-year average. 15 MR. KERR: Okay. 16 )UR. MINNERS: The three-year average, what they 17 resported is everybody -- I don't see~any-plants below 95. { 18 MR. KERR: So that would mean that Vogtle, cooper, 19'

Cook, Calvert Cliffs, and Zion either had this difficulty H2 0 '

after-1988 -- 21 MR. MINNERS: Or they didn't go below the E 22' reliability -- they could have gone below it in one year or L 23 at some time. Reliability is not a constant number. You 1^ ,j 24 don't hold it forever. It goes up and down. So this is a 25-three-year average. Yes. One year they might have had 100 l t I;

-t 32 .j l1 . percent,,and the next year they.might have had 94.- .,,.q ir tiL') ' 2 MR. CARROLL: But Vogtle hasn't been running that .?*' 3 long. .4 MR. MINNERS: You're talking about specific r 5 plants. 15 MR. CARROLL: Slide 19 is the history of 5fogtle. 7; MR. MINNERS: Yes, right. 8-MR. CARROLL: Seventy sensor failures. 9 MR. KERR: How long has Vogtle been running? ] 10 MR. CARROLL: I think about'three years. Does 11 that sound right.to you? 3 12 MR. MINNERS: Since about 1987. Because if you 13 read the IIT Report, there are pages and pages in it that -- 14 after the fact, when people went back and looked into the s 15 records. I don't know what the reliability -- we don't have 16 those numbers for Vogtle, do we?- We do and I'll get to 17 those., 18 This is just the data for the industry for one 19-year..So in many of these cases, we have plants -- back in-20 -1975, we had seven, twelve, fifteen plants below 95 percent, 21 and we had plants below. 22 So in every year here up through 1982 in this 23 slide, and continuing on -- 24 MR. KERR: Roughly, how many tests are normally . (. 25 made in a year? l i

-.,, ~ _ ~ -33 ,.[ -. 11 MR. MINNERS ( It's monthly testing on the diessis,. ~ ~ 2' so it depends on how many diesels you have in-your plant. s 3 MR. KERR:.But;it's'a' monthly test. 4 MR. MINNERS: It's a monthly test. 5 MR. KERR: Thank you.. 6 MR. ROSA: It could be more. .7 MR. MINNERS: The test.is-monthly,-but.you could 8 have more demands. Right. 9 MR. SERKIZ: You could be also-in accelerated 10 testing, because the old Reg Guide 1.108 Revision 2 had a Ell - count for;if you hit a -- what was it -- two out of 20, then .12 you went into accelerated testing. So if you go back in.the-M ) .13= history,- you'll find plants sometimes we were testing iL 14 weekly. 15 MR. KERR Okay. r 16' MR. SERKIZ: So you're at a minimum of roughly 12 17 per year. In some' cases, you might have 40 or 50 tosts on a 18 diesel in a year. 19 MR. KERR: Thank you. l 20 MR. LEWIS: With what cwnfidence, just looking at 21 those curves you just showed, the ones that are off to the-22 tail of the curve, say over.05 or something like that,-with-23 what confidence ir. the standard definition can one say that [(~} L24 that's real or as compared to a statistical fluctuation? %). 25 MR. MINNERS: I don't know.

@ DJ q o; .m y m -1 MR. LEWIS: In other words, is this just g 2 :- statistics or are those really bad plants?. f i

3 MR. SERKIZ:

Dr. Lewis, those plots that you see - 4 5 MR. MINNERS: This is just the unreliability as 6 just calculated using the test data. So you can look at 7 those other charts and what the probabiliEy is that that's a I, 8 false alarm. 9 MR. LEWIS:. I'm not talking about false alarm. 10 l'm talking about statistical fluctuations. We're talking 11 about a fairly small number of fa'ilures in all these cases. 12 ~ MR. MINNERS: Yes. 'l 13 MR. LEWIS: Therefore,'if an average is four or 14 something like that, there's a high probability that for 15 some places you'll get ten and-for others you'll-get none. 16 I'm just wondering whether these tail-ends of the curve 17 reflect that'or would they reflect real things I shouldn't 18 -be concerned about. 19 MR. SERKIZ: I don't recall what the number was, 1 20 but-these curves, which are Slides 12 and'13, were the 21 otaff's analysis that preceded the resolution of USI 44. 22 You can go back into that NUREG 1032 and that sort of 23 information is in there. I just don't recall what it is, 24 but I think Warren's point that he's making, also, is the 25 staff has conducted analyses in preparation of resolving A-

m 35-1= 44 and;you see that data on Slides 12 and 13. g.y S t - t,g - 2 MR. MINNERS: That's a two-edge sword, too, Dr. 3' Lewis. Yes,-there's a confidence that it could be better-l -4 athan that, but there is also a probability it could be worse 5 than that. 15' MR. LEWIS: I understand that. What I'm trying to i '7 find out is whether I'm looking at statistics or whether I'm 8 looking at something -- 9 MR SERKIZ: You're looking at histograms that m -10 have some statistical uncertainty with it, and I'dLhave to 11 look in the reportLto try to -- 12 MR. MINNERS: Ws were talking about it -- l ) 13 MR.-CARROLL: You also know, don't you, that 14 during that time' period, there were a lot-of real problems. 15 vith diesels. 11 6 MR MINNERS: Yes. 17 MR. CARROLL: That were worked on and fixed. 18 MR. MINNERS: Yes. 19 MR. CARROLL: That's what reflects the fact that 20 the reliability, even with the statistical uncertainties, is- .2 11 getting better. 22 MR. MINNERS: If you have a two-diesel plant, 23-that's only 25 tests and something like two or three f'T failures. That's right. That's not a lot of big numbers. -(x] _ 2 4 25 MR. LEWIS: That's right. Therefore, you should

i s 36 1 see it. l.C) L(- 2 MR. MINNERSt. The confidence is not 95-95, I don't t 3 think.- ll MR. LEWIS: I would guess it's zero.- No. 5 Certainly, as Jay said, there's other information besides. 6 these curves. 'I know that perfectly well. I'm just trying 7 to know whether I should take these curves as extra 8 information'or just go. entirely on this, 9 MR. SERKIZ:. I would cast it in the category it's 10 information that shows how plants slide in and out. It 11 gives you evidence of -- Slides 12 and 13 certainly show you y i I L 12 a picture that'says improvement is there, because you'd have k [ 13 less plantsLin the tail-out region, and we said that also in 14 a preceding elide, p 15 The.e has been improvement. Now, do you look at 16 an industry, average, and I think you should, and evidence on 17 an-industry average aggregate, whether'it's the-three years 18 that Warren had or otherwise, shows a good performance. } 19 The question also is, now, do you have an outlier 20 plant that is going to give you a significant safety problem L 21 that's lays n there in abeyance. We don't know. The reason 22 for the' reliability program, it is a program that gives you 23 a capability, whatever that is, to sustain what appears to ge-24 be a good reliability level. 'Q ) 25 MR. LEWIS: I don't disagree with anything you L l - - =

g PW %L II' 37 ') 1-said, except for the-part about.the outlier plant. I'm-p'(' 2 .trying'to' understand whether I'm'seeing plants sliding in .g a 3' and out of the acceptable band or whether I'm seeing ti, 4 statistics slide in and out that have nothing to do with the f: H' 5 plant. U,- 6 MR. SERKIZ: You're seeing plants slide in and out 4 W- '7 using these curves, not.the statistics that go with those U 8 curves. 9 MR. LEWIS: Well, you seem to be sure of that, and-10 I'm trying to find out why. 11 -MR. MINNERS:- You don't know. That's the problem I 12 with? statistics. You don't know what the truth is. It can 13 only try to give.you~some estimate-of'it. 14 MR. LEWIS: On'the contrary, statistics is a well-1 w 15 established, if you'll forgive me, mathematical science and 16 'there are measures of how confident you should be about what 17 you've learned from the numbers. -i 18 MR. MINNERS: I don't know if we have any numbers .19 that calculate that -- 20 MR. LEWIS: I guess I'm on a crusade against 21 either ignoring statistics or taking them too seriously, and 22 that I'm just pursuing the crusade here. I think this is a 23 particular subject in which it's important because you're 24-dealing with very small numbers of failures, and, of course, .q 25 you do want to assure the reliability, but you don't want to

w qq in 38 9 Il fool yourself with-statistics into taking actions that ib 21 really have nothing to do with reliability, o 3. I think that was partly Bill's question at the 4-very beginning of whether the reliability assurance program 5: is directed toward assuring reliability. I think that's i 6; what your question was in the beginning. I'm just being l 7' pushier about it. 8 MR. MINNERS: There are some plants which have 9_ like ten percent. You're pretty sure that that's not just 10 statistics. 11 MR. LEWIS: Well, when you say pretty sure, again, 1 r 12 that's something that can be quantified. There is certain (i 13 probability that those slide down below ten percent. I 14 looked at the numbers for Plant C and they do fit an 15 exponential distribution and, in fact, ten percent -- a 116 certain fraction of the time, ten percent of the plants will 17 fall below the ten percent level, o 18 I just wonder if the statistics are right. That's 19 all I want. 20 MR. SERKIZ: I don't have an answer for you on the 4 21 statistics on 1032, but I will look it up. 22 MR. LEWIS: Please go on. 23 [ Slide.] l ( 24 MR. MINNERS: We're on Slide 15 now. The last 25 slide that I showed you were industr'J data. This is some i V 1 - 2

m e-39 1-data that we had'our contractor do based on LER-data mostly, L2 and he has come up here, There's an eight-year history and 3 . here's a'four-year history in which one plant is just at the 4 95 level; another is just barely below it. L5 I'm not sure how much you can derive from this. 6-Let me say I think we-generally-agree with the industry's 17 statistics. (Slide.] l9-MR. MINNERS: This is a graph to illustrate that 10 the reliability goes up and down on these plants. This is-11 the number of demands between failures, so the target 12 reliability is somewhere in here. On this particular plant, ( 13 this is real data, we had good reliability and then it 14 dropped off, and-then we had better and it dropped off 15 again. =16 So you get these variations en( it's not a .17 constant thing. You don't get a reliability and it just 18 . stays there. It depends a lot on your. reliability progrr.n, 19 on how well you treat your diesels. 20 MR. LEWIS: Let me just, at the risk of which I've l ~ '21-already passed the test. That particular thing you showed l 22 does not demonstrate how that plant slides in and out of 23 reliability. That is an absolute statistical distribution 1 24 of an exponential reliability curve. It's completely I .25 . consistent with a uniform basic reliability across the

M@pa , y-q,, cl [..y y 40 g e ;3 e 1 entire curve in which.these are simply the fluctuating b QR l -fd 2 measurements. o n 3 104. MINNERS: But it's not a reliability that's up 4 at the 95 percent level. 5 MR. LEWIS: 'I'm sorry.- The one you just showed ^6-is. This one is not. '7: MR. MINNERS: I guess I haven't run the statistics ) 11!! 8 through. 9 MR. LEWIS: That's completely consistent with the-l 1 .10 uniform level, not a 95 percent. 11 MR. MINNERS: That's right. Not a 95 percent i L' l 12. level. _,$p -; -- 13 MR. LEWIS:- But the term you used, this shows how 14 a plant slides in andlout. It does not show that. It's a-i y : consistent distribution that you would-get from an .j ,l 16 exponential'with a uniform level, admittedly lower than 95 l 17 percent. Therefore, it's bad for that' reason, but not l L l r 18 because it's sliding in and out. 19-Now, the next one does show a secular trend. I' 20-agree. '21 MR. CARROLL: Plant B was.945. That's the same

. y 22 data.

i 23 MR. MINNERS: I don't'think this is the same l plant. l p. 24 W L 25 MR. LEWIS: The reason that happened --

i 41 J 1 MR.'MINNERS: I don't think this is the same' l r^Y ,j -(Q 2 plant. 3 MR. CARROLL: Plant B. 4 MR. MINNERS: Different reports. 5 MR. LEWIS: Also, with the thing like this, the 4 6 mean is dominated by the three high points, because that's 7 nearly 300 - This is a very funny way to plot the thing. j 8-MR..SERKIZ: But it does give an illustration, Dr. .9-Lewis, of even though the averaged or the apparent is a 10 constant or fairly level, and certainly within acceptable, 3 -11 that you do have these variations, to use your term. i 12-The variations in this case, if you go back and ! ( 13 'look at the data which was pulled out of tho LERs and NPRDS w 14 data, shows quite a bit of variability of underlying causes. 15-Mh. KERR: But the point that we're both trying to 16 make, and maybe we're completely wrong, is that the.. data-17-do not show'necessarily any change'in reliability. They may ~ ~18 simply be statistical fluctuations of data which-would occur ( 19 with a constant reliability. ~ 2 0. You people seem to be interpreting these as 21 indicating a change in reliability. l' 22-MR. MINNERS: You say these. This graph, I would-l. 23 . agree with Dr. Lewis, can'be a random variation. I don't 24 think the other ones are. (, 25 MR. KERR: I'm looking at 5-7, which you cra

42 4 la looking-at,_too. .2 MR. MINNERS: Yes. This can be interpreted as 3 random' variation around a rather low reliability. That 4' isn't 95-percent reliability. If you're going to have 95 5-percent reliability, you're going to have to have a tight 6 thing-in here. 7 MR. LEWIS:- Warren, you're reading it wrong, 8 because you're pointing your pointer on a log scale at the -9 median distribution. This graph represents approximately 10 300 demands with approximately 15 failures; 300 demands for 11 15 failures is 95 percent reliability on the mean. 'These 12 are fluctuations around that. You're putting your pointer '( 13 at the ten mark,-and that's wrong because it's a log plot. ~ 14 _You fool yourself_by putting your eye in the middle of that. 15' MR. CARROLL: So this does look like it's the same 16: Plant B. 17 MR. LEWIS: It probably is. 18 MR. SERKIZ: It is the same Plant B that was on 19 the table. That's correct. 20 MR. CARROLL: Approximately 95 -- -21 MR. SERKIZ: Approximately 95 percent averaged out .22 and sitting at that for the number of years shown there. '23 MR. LEWIS: These are the statistical fluctuations - l around that 95 percent. 24 25 MR. SERKIZ: These are the fluctuations, whether l

i 43- . 1_ 'you want to call them statistical, that's-the! result of i ) 2' where failures have' occurred over that timeframe, i s t 3 MR. LEWIS: Sure. No. I have no problem with. 4 that. + 5-MR. SERKIZ:.That's correct. 15 MR. KERR: But it is.not necessarily an indication

7 that the reliability changed during that period-i 8;

MR. LEWIS: It is not. 9 MR. KERR: That's -- 3 .10' MR. SERKIZ: It changed relative to a mean-and I - -i i -11 12 MR. KERR: I'm sorry. The data do not stume that, 1 13' really. It is possible that that occurred,'but from_these L.. 14 data, one cannot draw that conclusion. 15 MR. SERKIZ: From that data, you're correct. [ 1<6 Thera-is additional information in that report that shows 17 the variability. 18 MR. LEWIS: Well,_there may be information.- It's1 19-not additional information because these data do-not contain 20 that information. 21 MR. SERKIZ: There is information.in that report j 22 that you gentlemen are at liberty to look at. 1 \\ 23_ MR. KERR: Information about what? About numbers L 24 of failures? -25 MR. MINNERS: We would agree with you that that 'w 4

7; x, V m s w, 4 44 'l I f_ ' s y l 1l -previous graph here can be a' statistical variation. I think

I b

6 l =2 the' distributions that we showed:before are pretty much, I (~ r '3 would say wa ahow that there are some plants which are going. f4-to be outside the 95 percent reliability. h 3, 5-Those graphs show. plants with 90 percent a. h .6 reliability. ~Now, maybe you,should go home and do a ain 1

[

7-statistical calculation which says-what the confidence is, j f 8 but if the plant is down at 90 percent, it's pretty unlikely } ,m i + 9. that it's going to be really 95 -- 10 'MR. KERR: Warren, the point that we're trying to r 11 make is that those data do not show that the plants are down 12. at 90 percent. h 13 MR. MINNERS: I don't understand how you can say

u 14 that.

e 15 MR. KERR: I realize that you don't. t '3: 16 MR. MINNERS: I didn't say that. I said if a s 171 plant has.a measured reliability down at 95 percent, I'm ' 18 fairly confident that its reliability is less than 95.

Now, 19 I haven't done the statistical calculation and I can't prove

. 20-it to you in a rigorous mathematical way, but I think any .21 statistician -- when you do the calculation, you will find 22 that if you'get that far away from the target reliability, 23 you're not there. f 24 MR. LEWIS: I think you're wrong about that, and , k.)\\ 25 that's why I started this whole thing by saying have you i t

1 x !t 45 .g=h f([)I.2 fl -gotten professional statisticians -- if i'v - MR. MINNERS: Yes. 3 MR. LEWIS: I do not believe that they would, agree t 4 that' the' things that you've been saying about these. curves,-

5 and that's why I think it's extremely important,-if you're-6'

' going to inpose a program,-to do it right.- Doing things 7 mathematically right is not prohibited. j ,i J 8 MR'. MINNERS:' I guess we have a different 9 interpretation. o 10 MR. LEWIS: We do not. You don't know statistics. E ,i 11 -MR. MINNERS: That may be. 12 -(Slide.) .D( ) 113 MR. MINNERS: We do have more data.which shows a. 14 degrading trend on one plant. 15 MR. CARROLL: What does that show, again?. ' 16 MR. MINNERS: That there's a-degrading trendion i L x17 thin plant.; It's getting worse. 1 18 MR. LEWIS: Let's go.through that. 'I've looked at i this plant. This plant has approximately, on this curve, 17 19 20: failures and at1approximately 750 tries. That's 21L approximately a.975' reliability. 22-The last half of it, the lower part, should, 23-therefore -- has about 100 tries and, therefore, it should .24 have, a t ~.two-and-a-half percent, it should have had maybe t [f .25 three failures. Instead it had seven.

w 46- .1 Now,'I' agree that if you expect three and see ^ /"'s , 'ks-1 2 seven, you should begin-to think about it. But on the other 3 hand, it does not. demonstrate with any. confidence, according 4 to'a real statistician, that you've seen a degrading trend. 5 You may have. I. don't think it's excluded, but I don't 6 think it's demonstrated either. 7

You' eyeball these' log normal plots and you can 8

come to seriously erroneous conclusions. You have to do a -9 real statistical analysis, mathematical, to find out with. 10 what confidence you can say this is a declining trend. It's i Ell: three'against seven. I wouldn't bet my general welfare on 12 that. N( ) 13 MR. SERKIZ:.These types of discussions-were gone 14 through, though,LDr.. Lewis, and to settle out on what -- in- -15 your Slide 6, what.we term trigger val'ues -- the type of. 16 information that was utilized to'come up with what I'll term' 17 a simplistic' set of trigger values to try'to take that into 18 account. 19 As the Slide 6 shows, there are certain actions 20 , thatiare set-up to carry through in a : reliability program or 21 a maintenance program, and-they become more detailed as you H22 move from left to right. 23 The statistics that go behind those numbers, that 24 information is available, can be provided. Again, I think' [ 25 the point I would want to make and we are making is not

s@ [ 47 ' tii:+ 1 using statistics to drive'this,.we're trying to leave a set 2 of operations that is effective,-in place,-and provide 4 3 guidance that's common to both the staffer and the industry, 4 -to leave.well enough alone, but at least'have a common

5 ground to discuss what actions do constitute a reasonable

+ 6 approach-to reliability programs. i 7-MR. LEWIS: I've a lot of sympathy for your l 8 problem. I'really do. And I want to essure reliability, 9 too, but I don't want to point after it in directions in 10 which you fool yourself. 11 The problem-with trigger levels is actually an { 12 extremely interesting one, because if you're trying to-

(

13 detect trends in a moving situation, people'do sometimes-use 1 14 trigger values. I could find thousands of books written on 15 -how to make money in-the stock. exchange by noticing whether 16 the Dow Joned has gone up three times before it went down 17 two times, or whether it went up three percent -- trigger 18 levels. People look for triggers. 19 It's a human failing that we look for triggers 20 that tell us yhat to do. Some people went to oracles. They 21 don't always work. I'm only asking that this be done 22 professionally. I'm not asking anything else. 23 MR, SERKIZ: I can assure you that the proper 1 24-professionals were brought in and the approach on trigger ,( 25 values is an approach that both the industry and we felt was

1 j!, ,a .j s' r p x by ) 48 ? J i s, ?>- algoodLway to look at the problem,-to look for trending or il y(V L2. ' 5 4: lwhatever way you want to continue that discussion. ] (Slide.) 3

4; MR. MINNERS

I want to skip over some of the 5: _ slides there, which just give you some examples of 6. = maintenance. problems with diesels. That's what we're 'l P 7

talking about.

That's the central point of the discussion. 8 Is there'a' problem? Do you need something besides a 5L monitoring program to assure diesel generator reliability? 10 I think that's our basic difference with NUMARC. r ill; They're saying you need a program, but you don't need it l 12 documented in a Regulatory 3uide that specific; that people l' M {( )! 13' .are taking care of their problems and-they don't need . c 14 anymore -- 15-MR. KERR: There was a rather extensive discussion 16 of a maintenance rule that took place over the past two, 17 two-and-a-half, three years. If my memory serves me 18 correctly, the Commission decided, after some deliberation, 19 not to try to adopt a maintenance rule. 20' MR. MINNERS: At this time. 21 MR. KERR: At this time is what I'm talking about. 4l 22 It appears to me that the staff is advocating a de facto 23 maintenance rule for diesels. (' .24 MR. MINNERS: Could be. O} 25 MR. KERR: Which the Commission either knows about

4 ~1 49 l' and approves and, therefore, goes contrary to the decision 7-q _ they made earlier, or which the Commission doesn't know 2 3 about yet. 4 MR. MINNERS: The commission reviewed and approved 5 the Station Blackout Rule, which is the basis for the 6 reliability program on diesels. It was clearly explained to 7 the Commission at the time of the Station Blackout Rule that 8 one of the elements of this was we were going to later on 9 specify the detailed guidance for diesel generator 4 10 reliability programs. i 11 MR. KERR: But a diesel generator reliability 12 program could be one which said you take data to determine L ?~' i, ((j% '13 reliability. It doesn't have to tell how -- 14 MR. MINNERS: We had the elements of a diesel 15 generator reliability program and:it wasn't just monitoring. I 16 That was specified at the time. We didn't have the details 17 of it, but there could be no misunderstanding that we meant 18 more than monitoring. 19 MR. CARROLL: When you say "at the time", was "at H 20 the time" at the same-time the commission was thinking more 21-seriously of a maintenance rule, though? 22 MR. MINNERS: I guess the Station Blackout Rule 23 came before the maintenance rule, but I think they were } thinking about it still then. It wasn't that long ago. 24 25 MR. SERKIZ: The subject of a diesel reliability

ll

~

50 L I\\ < l~ Program was very actively discussed.- It's also noticed in a y~ q(- Federal' Register Notice that went out on 844 that there i

2-al--

-t 3 would be a' reliability program that would come out of B-56. (Y 4 The Reg Guide 1155 han, in a general sense, some very 5 similar language to these elements. 6 The intent here was to develop the guidance that 7 would supplement the Reg Guide 1155 and to establish a 8 common set of guidance that both the staff and the industry 9 .would use. 10 MR. MINNERS: The staff, as always, is at a 11 1 disadvantage that we don't have the data on diesel generator a L 12- . failures. We cannot get that data. We don't have it. 'So i l' l()113 tocmake these' statistical calculations that you're 14 proposing, we can't do that at this time. 15 lo get that data, we would'have to send out a 16 letter to the industry asking them-for the information on 17 which to calculate that data, and I find that that's always ' 18. -a difficult position that the staff is in. 19 MR. KERR: Diesel generator failures aren't 20. reported to the staff. 21. MR. MINNERS: Only half of the plants have a -22 diesel generator reporting requirement. 23 MR. SERKIZ: The majority of the reports that ccme /~} 24 in.through LERs on diesel generators are tied to what's g-25 called, I guess in my terminology, special reports that are

P 51 j 1 tiediinto plant'tachnical specifications. l t yg- \\s,f 2 Those plant technical apacifications, that have i ii 3 embadded in them a compliance with Reg Guide 1108 Revision L 41 2, then report these fa'ilures. Now, if you take a look at 5 1the.90 or 100 plants out there, approximately half have -l 6 technical specifications and, as a result, they report every t t 7 failure. The other half, roughly speaking, have no such i 8 reporting. requirements and do not report them unless they 9 come up for a different requirement. ~i ~i 10 MR. 'KERR: So if you had these data, however, you 11 still wouldn't be assured that you were achieving the q 12 reliability you want. You would still want this additional IIY 13 assurance that will be provided by the reliability program -X s! - \\ 14 to which you refer,

15-MR. MINNERS: ' lie think if we had the data, it 16 would show that there are plants that are -- you would make 17 your best estimate that the plants are below 95 percent

-18 ' reliability. People go that way. And that's based 19 partially on what we see from the statistics that we do 20-have, plus our experience with what happens out in plants. 21-They do have trouble with their diesels. 22' MR. KERR: Well, the statistics which you do have 23-most recently seem to show that the plants are all achieving r 24 at least 95 percent reliability. So what -- 25 MR. MINNERS: No, sir. There isn't one chart I

52; 1 showed up-there which showed that all plants were above 95.- 2 MR. LEWIS: All'wasn't the issue. .3 MR. MINNERS: That's what he said, all. 4 MR. KERR: I thought you -- 5 MR. MINNERS:: I would agree with you that most 6 plants are above 95 percent reliability. As I said, I don't 7. think, in general, that there's a problem with diesel 8 generator reliability. It probably is very high for most 9 plants. 10 We're always in the situation of' regulating the 11 bottom ten percent of the plants, and I think there is a 12 bottom ten percent. 13~ .MR.. LEWIS: That's what I've been trying to find 14 out, whether I should believe there is a bottom. ten percent 15 or -- 16' MR. MINNERS: That's a. good question to ask, and 17 this is the data that we have and we don't have.any more 18-data. To get more data, it would take a very large effort- -19 and I think even the industry would probably object to us 20 sending out a generic letter saying send us the data. 21 MR. KERR: I-thought the reporting requirements 22 that were going to be part of this Reg Guide would give you 23 the data. Is that not the case? ~ 24 MR. MINNERS: No, sir. !e 25 MR. SERKIZ: That's not the case.

' 53 - l 1-MR.-MINNERS: No, sir. What I've tried to say is 2 that we want to get out of the diesel generator reliability. 3 business. This, Reg. Guide would turn over diesel generator l 4 . reliability' completely to the industry. They would do the L 'S monitoring.'.They would do the maintenance. They would do-i 6 the'whole thing. 7-Only in-the cases in which the generator 8 reliability gets really what we think is very bad, what.we- ~ 9 call,a problem diesel, would they have to report'to.us.. l~ 10

MR. KERR:

I see. They'd have to report that. 1L1 MR. MINNERS: So we don't have any reports coming 12 in,.and NRR doesn't like that. I can kind of sympathize l( 13 with them. But:we have said that we don't need to: monitor '14. diesel generator -- the NRC doesn't need to monitor diesel l -15 -generator reliability as part of its reasonable assurance to 16-the public because we've got such a good program that the. 17-utility is implementing. That's our basic argument. .18 MR. SERKIZ: The only reporting that is required, i 19-will be required under the current Reg Guide, is a reporting 20. of-the problem diesel. The recordkeeping.which is in-21 Section C.4 is the same as the recordkeeping outlined in the 22 NUMARC Appendix D. The Reg Guide -- 23 MR. KERR: That would be something new? 24 MR. SERKIZ: No. The recordkeeping, in my /%( V 25 understanding, is consistent with most plants, whatever that

1 54 1 -1 means, because we're?getting back to what most plants is or OV 2 are. 3 MR. KERR: But that's just on-site. 4 MR. SERKIZ: And that is kept on-site. The -5 monitoring is done and maintained on-site. The special 6 report that we're asking for in;this Regulatory Guide would J 7 be for the problem diesel. .8L Our point of view is that if you get into a j i 9 situation.that-both the industry and we have termed the 10 problem diesel, that comes up with a four-out-of-25 count, 11 the NRC should be notified and certain information provided. 12 Right now, with the Reg Guide 1108 Revision 2, half the .(O) 13 plants' report every failure. 14' MR. WYLIE: We're going to have to move along. 15' .MR.cMINNERS: -I can get quickly through this. 16' ~ (S1ide.) 17 MR. MINNERS: This is where we are at'this point. 18 =We'have this difference. We've gone to CRGR and based'on 19' their recommendation, we're going to modify our Reg Guide, 20 as I said!before, to take out positions C-2 through 7 and 21 put those -- is that right? 22 MR. SERKIE: C-6.2 through 6.7. 23 MR. MINNEFS: C-6.2 through 6.7, pardon me, and 24 put those in an appendix with a statement that the material 7 t in the appendix is just illustrative and gives examples of 25

/ l 1 c -p n - i ,,,s'pp-1.1 llL

t&

f "g ,x l.[( ' y1 1-things that you could do, but is not part of the staff's f K2i position, in order to try to give utilities as'much l

n

(

3-flexibility as possible.

C b 41 [ Slide.) 5 MR. MINNERS: So that's all that this slide says. '6 .We have some slight modifications to the 5054 letter to try 7 to say what we want to say. We always had problems with-7 -i 8-those letters and getting them to say the right thing. We 9 do intend to make a formal reply to the comments that NUMARC 10 sent to us on the Reg Guide to explain our positions on each - 11 one of those comments. 12 (Slide.] n kj - 13 MR.-MINNERS: So what we are planning to do is to 14 issue the Reg Guide, as you see and, as I told you, we're-l 15 going to modify it, to issue a 50.54(f) letter that would 16 get licensees to react to this. 17 MR. KERR: So if we wanted to know what was going 18 to be in the Reg Guide as you finally plan to issue it, we'd a 19 take what we have in this material and move C-6.2 to 6.7 to 20 an appendix? +, > 21L MR. MINNERS: To an appendix, and it will be [ 22 clearly noted because of all the concern expressed, that ti-23 this there for illustrative purposes, t.'e tables are f'24 labelled examples, and so forth. I believe that the 25~ material in those sections is consistent with earlier

ff.s m '56i 11. versions of'the; Appendix =D. (T 1 t N.. 2 =The illustrative material or tables are direct i .3 extractions and we'll put it in an appendix if it gives 4 people that much of a pain. 5 MR. MINNERS: That would close out the issue-for 6 us. We're going to issue, par our standard procedures, a 7. Federal Register Notice announcing-the availability of the-8 Reg Guide,'and.that will contain a backfit analysis, which i '9 really refers back to the 844 backfit analysis. t 1 10 MR. LEWIS: Just a matter of curiosity, I don't 11 know the rules. Do yi.u have to issue a backfit analysis 12 when you issue a Reg Guide which in not mandatory? J 13 MR.-MINNERS: Yes. -i 14 NR. LEWIS: You do'? i 15-MR. 71 INNERS: Yes, sir. \\ 16. MR. LEWIS:' I didn't know that. l 17 MR. MINNERS: ' If.the staff has a new 18 interpretation of their position, we have to'do a backfit -19 analysis. That's what the --- i 20 MR. LEWIS: I just didn't know that. 21 MR. MINNERS: 109 is stated in terms of position, 22 not.in-terms of regulations. The staff had a position and p 23 they're changing the position. You have to do a backfit L 24 rule. L 25 MR. LEWIS: That's right. You're right. It i i.

57 1- -1 refers'to branch technical positions and things like that. /( i i 2 MR. MINNERS: 'It doesn't talk about regulabons. l '3-MR. WYLIE: Warren, as a matter of clarification, O 4 though, now, you are now proposing to move 6.2 through 6.7 5-into an appendix as written or are you going to make any 6 modifications? 7 MR. SERKIZ: No, its written. What you have in 8-those sections will be moved to an appendix, and-it will be q E 9 noted now in Section C-6, which is the introduction part of 10 it, that they're being put there for -- -i 11 MR. MINNERS:- In the latest letter to the 12 committee that we sent down, we put in an example of how we L 13 think wefre going to do that. So there are some words there L 14 you can read. Al, do you have'an extra copies of that you 15 can hand them right now? i 16 MR. SERKIZ: I think so. 17-MR. MINNERS: Because that is kind of a last-1 18 minute change. Dr. Lewis, you're correct. The question is 19-is there enough of a problem; that is that there are plants l-20 that are below the target reliability; that we need to have L. 21 something beyond the monitoring program. 22 MR. LEWIS: I'm prepared to believe that there is 23, a problem on the basis of other things than the statistics, 24 but I'm not convinced by the statistical treatment at all. 25-MR. MINNERS: I wish we had brought better

58 1 statistics and I apologize that they weren't -- (,- V 2 MR. LEWIS: Don't apologize. We're in this 3 together. But I do think that it would be important to -- 4 MR. MINNERS: As I say, we don't have the data to [ be able to do what you asked. 6 MR. XERR You have reason to believe that the 7 problem plants are not using these preferred practices that 8 the industry has concluded are good pract. ices. 9 MR. MINNERS: Those set of slides I didn't show 10 you which gives you all the problem areas, when you raad 11 that, it says things like people don't prioritize their 12 maintenance or.they didn't do trending, or all of the things l 13 that we would think would be in a good reliability program. 14 MR. KERR: And they nav.J't learned yet, after 15 having all that problem, all those problems. 16 MR. MINNERS: Maybe the guy who learned went to 17 work for another utility and they got a different guy in 18 there, or maybe they got a new manager who caja, hey, I'm l l 19 cutting costs and we're not doing this. People's confidence l l 20 goes up and down. Or maybe the guy had a fight with his 21 wife the night before and decided he wasn't going to work l 22 very hard that day. There are lots of reasons that things 23 degrade. 24 MR. KERR: When you get out of the diesel 25. reliability business by formulating the Reg Guide, you won't l

59 i c 1-have'any way of knowing whether they're following the Reg ,,, ~ 2 Guide or not. 3 MR. MINNERSt That's right. We don't intend -- 4 MR. KERR What assurance does that give me that 5 things have improved? 5 6 MR. MINNERS Improved? I think what we're trying ? 7 to do is maintain w.lat we have. We don't think what we have 8 is bad. We're just trying to maintain what we have. We 1 9 think diesel generator reliability is pretty good. 10 MR. KERR But if you don't have any way of t 11 knowina whether people are following this Reg Guide -- 12 MR. SERKIZ: Yes. The 50.54 (f) letter is (( 13 currently worded, it says are you going to follow section C-14 3 -- 15 MR. KERR: But you say you are going to get out of 16 the reliability. To me, this means you are.'t going to 17 inspect and see whether people -- 18 MR. MINNERS: That doesn't -- first of all, if he 19 gets a problem diesel and starts reporting to it, that's 20 going to trigger us and I'm sure there is going to be some 21 inspection about the people who have problem diesels. 22 Also, resident inspectors can inspect anything 23 they want to anytime they want to. 24 MR. KERR: You mean they're not part of the NRC, 25 then, because they're not going to get out of the diesel l 1

. _ ~. ,. 9 60 1 reliability business. [] V 2 MR. MINNERS: But that's just their general j i 3 auditing of licensee performance. When I say get into the 4 diesel reliability, I mean that on a day-to-day basis that 5 we'd get the reports and monitor it. No. We have an j 6 auditing A nction to go and look at plants just because -- 7 it's just like training or anything else. i 8 tie've turned training over to the industry. That 9 doesn't mean wa don't ga out and audit how they're training i i 10 is doing, but it'. certainly turned over to the industry. 11 We're not in the day-to-day business of monitoring training. t 12 MR. CARROLL: or prescribing how it's done. l 13 MR. MINNERS: or prescribing how it's done. That 14 doesn't mean we completely give up our responsibility to do 15 an audit, to be able to go -- we have to be able to say to 1 16 the public, hey, we've checked once in a while to keep 17 people honest. l 18 But what level that will be, I don't know. That's 19 up to the inspection people. 20 MR. CARROLL: I would pick different words than 21 the ones you've been using, getting out of the reliability - 22 23 MR. MINNERS: I think maybe that's true, but in 24 the traditional sense of how we used to do things where we 25 were on top of people every day, that's what I mean by being ~

61 { t 1 in the business. We can't obviously completely absent t ,es 4 V 2 ourselves from plants, but I mean that we're not going to do I 3 it on a day-to-day or a periodic basis. ) 4 MR. SERKIZ: Let me make a clarification point, 5 because the a m ms to be a going back and forth. There's a 6 monitoring **' ton of this, which is Section C-3 of the 7 Regulatory Guide. The monitoring portion in the same I 8 approach, the same treatment values, etcetera, as in the 9 NUMARC Appendix D. 10 The reliability program elements, where, if you 11 were going to have to go into an inspection type role, are 22 contained in Section C-6 and now supplemental information -( 13 will be moved into an appendix. 14 The monitoring will continue and that will be done 35 in the field. As one of the slides that Warren showed you 16 indicates, we rely in total on the industry on telling us on 17 the health, operation.and so on of the diesels. We are 18 relying on the INPO reports that come through on plant l 19 performance data. 20 You have a Slide 12 -- I'm sorry -- a slide 14 21 that shows you annual distributions. If you look into a 22 more recent report on the year that follows, you will see a 23 larger number of plants out to the right. D( 24 MR. MINNERS: INPO has a monitoring program that 25 they're doing all the time. They're monitoring diesel

i i 62 1 generator reliability industry-wide. _2 MR. SERKIZt So for us to get in and do an s-3 independent analysis at this point and go back, thr.sbvious 4 rebuttal or reply would be, well, you're impacting es; we've 5 already done it; the records are there, etcetera. 6 The data we're presenting you is a twofold set of [ t 7 data. We showed you what data was assembled by the staff in 8 resolving A44, and that's there for illustrative purposes to 9 show you that it's been an improvement. 10 We showed you some industry data that was made 11 available to us as part of meetings with the NUMARC B-56 i 12 Working Group and it shows, on the industry average, v ry If ) 13 good performance, and it shows also in slide 14 a sliding in 14 and out. 15 The point I'm trying to make is monitoring the 16 performance through Section C-3 or NUMARC's Appendix D is an 17 ongoing function, and that will be done, and is consistent 18 with the definitions used by INPO, etcetera. The 19 reliability program, which people like to talk about and i 20 talk about good practices, etcetera, we've tried to set down 21 in print. 22 We had side-by-side documents set up in November I 23 and December. The CRGR suggested that we junt adopt in /~% 24 total, by reference, that Appendix D of NUMARC. We said i 25 fine. Discussions were held with NUMARC. You have the

63 i i document that was submitted by NUMARC. The illustrative p(Nl 2 sxamples, the detailed guidance, etcetera, was taken out and I m-3 put into a Topical Report. 4 It is the staff's position that having the general 5 guidance we have is not overly prescriptive, but if you have 6 none of it, then we will have the situation where you have 7 inspectors going out or staff going out and talking from a i i 8 different cookbook. 9 MR. WYLIE Thank you very much. Let's move l l 10 along. We'll call on NUMARC now to give a presentation on 11 how they see things. 12 MR. RAISEN I'm Bill Raisen, Director of the + l l I()'13 Technical Division of NUMARC. I'd like to mak4 a few 14 comments, and then Alex Marion is with me who has managed 15 this whole interaction with the staff. We have provided you l 16 with, by way of letter to Charlie Wylie, our letter to CRGR 17 containing our problems and objections to the staff's 18 proposed generic letter and Reg Guide. l 19 I would like to make a few introductory comments t 20 because what we are really disagreeing with here, I believe, 21 is regulatory impact and regulatory philosophy. As Warren i l 22 told you, we had a lot of interactions with the staff over a 23 long time, and our people used to come back with stars in 24 their eyes over these statistical arguments. 25 But I think a very good consensus was attained-1- 1. L i - - - - ~

L i 4,5 64 \\ + 1 there. In fact, we do not have disagreements with that, nor ') i 2 do we have disagreements with guidance on what kind of 3 things have been successful in a maintenance procrca or a 4 preventive maintenance program. 5 But, in fact, we do have some serious problems i 6 with the manner of implementation and the regulatory impact. 7 In fact, there was a lot of talk about a Reg Guide isn't a ) 1 8 requirement and, in fact, it's not. But if you look at the 9 practical path, in fact, the 50.54 (f) letter asked the 10 utilities to make a response that, in fact, committed to 11 certain sections of that Reg Guide on their docket, which, 12 in fact, makes that an enforceable requirement on the fO 13 utility. g 14 In fact, that would be the situation. Now, we i 15 agree with the diesel generator reliability needs as stated i 16 in the SBO Rule, and, in fact, NUMARC specifically asked 17 each utility to make a supplementary response to their SBO 18 submittals and specifically requested that they indicate in 19 that response on their docket that they understood they were 20 required to maintain the reliability chosen in their SBO 21 response, either.95 or.975. L 22 So we certainly have no objections to that, nor no -23 objections to that as a regulatory commitment. I think we ,( do have a little bit of a difference of opinion with the 24 25 staff on the definition of a reliability program.

65 1 In fact, NUMARC, through a formal NUMARC 7 l 2 initiative, binding on all of the utilities, committed to 3 the target values that are in our Appendix D and consistent 4 with the staff's Regulatory Guide, and, in fact, the action 5 chart that Warren showed you and Alex will show you as to 6 what you do as you hit any one of those triggers. 7 We fully support that as a reliability program and 8 have no problems with it. The staff is right to be somcwhat 9 concerned with us, I think, for taking the detailed " good 10 practices" kind of material out of Appendix D. l 11 In fact, we did that because in discussing the 12 impact and looking at the staff's Reg Guide and in looking (( ) 13 at the results of our discussions, we felt what would happen i 14 was what we saw in that draft generic letter; that the f i 15 industry was asked to commit to Section C-6 of the Reg 16 Guide, which, in fact, included a lot of detailed preventive e 17 maintenance and maintenance tasks which were people's 18 opinions on what might be good things. I 19 We don't disagree they might be good things, but r 20 making a regulatory commitment to them, we thought, was not 21 proper. So we have objected to the process in that regard. l 22 MR. KERR: Now, is it your understaading that 23 these things would be inspected against, and if they weren't q[ }. 24 being followed in detail, the plant could be cited even 25 though the reliability of the diesels was quite acceptable?

1 I 1 66 i 1 MR. RAISEN: That's correct. Moreover, if you ( 2 looked in the Reg Guide, there was a requirement in there -- 3 the tables give mora examples. However, there was a j l 4 statament in the Reg Guide which the utility would have had 5 to commit to that says, well, these are examples; you don't 6 have to do it exactly like this; however, you have to give 7 us an analysis of your maintenance program showing us that i 8 it's equivalent to this. 9 We felt that even that task was unwarranted and 10 uncalled for. Those are the philosophical bases for our 11 objections. We do hope we can continue to work with the 12 staff, because we, in fact, do agree that the direction (('T 13 we're headed is a good way to do business. v 14 I'm not sure -- 15 MR. KERR: What is the direction in which you are i 16 headed? 17 MR. RAISEN: With the rollability program and 18 agreed-upon trigger values as indicators and agreed-upon 19 actions that one takes when those trigger values are hit. 20 so that you do maintain reliability on an ongoing basis and 21 -you do assess your prograr, in total, if there is indication 22 that, in fact, your reliability is slipping. 23 MR. KERR: Okay. Then the problem you have is 24 that of committing to the details of a reliability program O 25 as part of one's license or --

.~. .= i 67 1 MR. RAISEN: Well, it depends on how you define g- ' ~ ' 2 reliability program. I'm not ready to define a reliability ] 3 program as having to have a detailed cocm!* red on the docket 4 maintenance program. 5 MR. KERR You tell me, then, what it is -- I 6 still' don't quite understand where you and the staff part 7 company. 8 MR. RAISEN: We part company on a docketed i 9 commitment to the Reg Guide that is an enforceable action on I 10 the industry, and particularly to Section C-6, which 11 contains all kinds of details as to what your PMs and 12 surveillances and whatnot should be on your diesels. d 13 MR. KERR: Isn't this what the staff calls a 14 reliability program? L 15 MR. RAISEN: From Warren's talk, I would say yes. i 16 They include -- let_me say again we don't say that you don't i 17. have to do maintenance on a diesel generator to have good i 18 reliability. That is foolish, obviously, and an element of l overall good reliability is doing good maintenance. If 19 p L 20 you're not doing good maintenance'over the long run, you're L l 21-probably not going to have good reliability. 22 That doesn't mean, however, that you have to have r 23 a commitment to some table of, well, I do these PMs on a 24 daily or an every-shift basis, or justify why you don't do j 25 them on a daily every-shift basis.

68 1 so it is really a regulatory impact kind of 2 concern, and I agree with the statement that, in fact, it 3 was de facto maintenance rule. That doesn't mean we're 4 against diesel generator maintenance. 5 I think that we h3ve to be careful that we 6 understand what the difference is between us. There's been 7 a lot of good technical work go on and I don't want to call ) 8 that good work into question. 9 I wonder whether we're all ready for this kind of i lo approach. .i think even Warren's presentation was rampant 11 with examples of, well, gee, even if you have 95 pe'* cent ) 12 reliability, sometimes you have failures. You do. That's a () 13 fact of life and we're always going to have failures, and 14 you're always going to have opinions on whether maintenance ? 15 practices are good or not. 16 If you look in Warren's slide, he looks at Calvert 17 Cliffs and has some diesel -- I guess those are reliability 18 numbers, but then an indication in the inspection reports 19 that they didn't like their maintenance program. We'll have 20 differences of opinion. 21 The NRC will continue to inspect this area, and we 2? welcome that, and they certainly should. But that's 23 different than a dockened commitment to a detailed, 24 essentially, maintenance program. 25 So that's where we really come down towards -- ,---r,..

69 1 MR. KERR In the slides that Warren did not show ( 2 in detail, there are mentions, some rather egregious 3 examples, it seems to me, of failure to achieve diesel 4 reliability. What -- 5 MR. RAISEN: It's not clear to me that that's true 6 from his slides. There are some egregious examples, I 7 think, perhaps of independ6nt failures or maybe even series 8 of failures, but I don't see the correlation to overall 9 reliability. 10 MR. KERR: You would say that those plants did 11 have adequate diesel reliability? 12 MR. RAISEN: From the INPO data since 1982, yes, 13 and according to the program and the trigger values that we 14 have set up. That doesn't mean that you aren't going to i 15 have egregious examples of bad. failures or perhaps bad 16 maintenance that leads to the next failure that you have to 17 do over again, i 18 MR. KERR: But from your point of view, these 19 diesel reliabilities from this -- from your knowledge of 20 diesel reliability generally, you think the diesel 21 reliability at this plants was appropriate. 22 MR. RAISEN: I'm not willing to speak to 23 particular plants on incidents which -- I don't know much 24 more about these than are on the slide, except in maybe one 25-or two --

) 'a i 70 ll 1 MR. KERR I'm not talking about what's on the 2 slide. I'm talking about the reliability of the diesels at 3 those plants. 4 MR. RAISEN: I am confident that the technical 5 work done, the trigger values developed and the actions 6 behind those trigger values, is an adequate mechanism to 7 take care of these. If, in fact, a utility has a program 8 that is bad and is not maintaining the diesels well, you 9 will hit the triggers. I 10 If you hit a double trigger, you go into a column 11 on that response chart that it requires you to do a 12 programmatic review, not only of your diesels, but of your 13 programs for maintaining those diesels, and I think that's l 14 entirely appropriate. 15 I don't thiak it's an indication of any one 1 16 particular incident here -- 17 MR. KERR: If they'd lieen doing a lousy job of 18 maintsnancs, what makes you think that they'll do a better 19 job of analys.is and suddenly correct their maintenance just f 20 because they hit a trigger value? 21 MR. RAISEN: Perhaps they won't, but I'm also 22 confident that the staff is really not going to disappear, 23 as Warren may have indicated and I'm sure he didn't mean. 24 The values, the failures of the problem EDG have to be j/ \\ 'd 25 reported and I believe the document requires hitting the

71 'l trigger values to be reported to the staff. 2 I am confioc.7t that they are not going to ignore 3 that. If a utility's corrective action is not adequate, the 4 staff has all the regulatory authority they need to pursue l 5 that, including the docketed commitment of the utility that 6 they would maintain the reliability that they claimed in .I 7 their Station Blackout submittal, either.95 or.975. 1 8 We have great confidence that the staff will 9 exercise their regulatory responsibilities. 10 With that, I will let Alex come up and give you a 11 very quick look at our response and let you ask any detailed 12 questions you might have. I see no need to belabor these (( ) 13 points, unless you have specific questions. 14 MR. MARION: Good morning. My name is Alex Marion -15 and I'm with NUMARC. What I would like to do is overview 16 some of our activities that have gone on for the past couple 17 years in working with the staff to develop a resolution of 18 this~1ssua. 19 The purpose of the NUMARC involvement on this was 20 twofold. One was to develop a reliability program.

Now, 21 because of the fact that we have an obvious difference of 22 interpretation of what the program means, let me just refer 23 to it as a process, but a process for maintaining and 24 monitoring the chosen station blackout target reliabilities

((g~g 25' of.95 and.975. .~

l 72 1 As Bill indicated, the utilities have already (,n \\m / 2 docketed a commitment to maintain those chosen targets. The 3-second purpose of the effort was to provide a basis for t 4 resolution of Generic Issue B-56. As a result of our 5 activities to date, we believe that there are three aspects a 6 or three elements that, together, provide closure to this 7 generic issues. 8 The first is that the intended goal of B-56 was to 9 achieve diesel generator reliability to a level of.95, and 10 that it was also an intended goal of the Station Blackout 11 Rule. And industry performance has not only achieved this i 12 goal, but has surpassed it. I f( ) 13 - The INPO plant performance data and the EPRI data 14 that was published in NSAC-108 indicates the reliability has 15 been greater than.98 since 1983. In Warren's presentation, [ t i l 16 he showed you chart similar to this that showed data on 17 unreliability through 1988. 18 So if you pull that out, you'll see a similarity i 19 of this through this point in time. This also has the 20 average unreliability data for 1989, which is indicated 21 here. On the righthand side, I've indicated what the 22 reliability targets are that utilities have chosen and how 23 those targets relate to the average-industry performance. gr g Now, this data that was provided via EPRI and INPO 24 25-or the program that's managed by INPO has been accepted by l L

73 1 the NRC. I believe at the last nesting we had, the NRC 2 indicated acceptance of the EPRI report and the NRC has also 3 indicated acceptance of the plant performance indicator i 4 program -- l 5 MR. KERR Excuse me just a minute. The statement 6 was made by the staff that they had no way of getting these i i 7 data, so they didn't know what was happen'ing at all of the 8 plants. Only about half the plants provided them data. How 9 do you get these data? 10 MR. MARION: This is data that was provided to us-11 by INPO and INPO publishes quarterly reports on the plant 12 performance indicator program. I believe a copy is sent to k 13 AEOD. At least one copy goes to AEOD, and this is on a i 14 quarterly basis. 15 But the purpose of the plant performance indicator 16 program is to collect unavailability data, and that 17 information is provided. And one of the areas that is 18 looked at by that data collection process is AC power system 19 unavailability and diesel generator performance is captured 20 in that. 21 MR. KERR: That wasn't what you were talking 22 about, then, when you said that you didn't have data for at 23 least half the plants? 24 MR. SERKIZ: The data I was referring to is data 25 that has come through because of current technical

74 t i specification requirements. Roughly half the plants submit .rs k-2 a report. The data we're referring to that we don't have 3 the detailed information is the information that's referred 4 to here as the INPO plant performance data. -5 MR. KERR: And you don't get that? 6 MR. SERKIZ: That's correct, sir. I 7 MR. CARROLL: On an individual plant basis. 8 MR. MINNBRS: We get the summary data that he 9 showed here. That comes in the plant performance report. 10 We don't get the data on which that is based, and you would 11 need that data to make calculations such as Dr. Lewis was-i 12 talking about, is what is the confidence and all that kind () 13 of stuff. p' p 14 MR. KERR: But somebody has those data. INPO gets 4 15 them, apparently. Otherwise, they couldn't assemble these. 16 It's just that they don't make the individual plant data 17 available to NRC, P 18 MR, SERKIZ: Correct. 19-MR. CARROLL: Which leads to my question. I guess 20 we're all familiar with the story about the guy who drowned L L - 21 trying to get across r river that only averaged three feet l L 22 deep. What's the band on these numbers? How deep is the L L 23 011? Are there plants in 1989 whose reliability was poorer 24 than -- 25 MR. MARION: There is one plant that is an l l l 1 m+ +, n + ~ w

75 1 outlier, if you will, for 1989. Now, if you look at the p i Q 2 performance of that plant over a three-year period, you will 3 find that the plant is not an outlier. 4 Now, what happened with that particular plant was f 5 the fact that they upgraded their control system for the 6 diesel and they had a number of failures over a short period 7 of time. And it's the number of failures over a 50 demand 8 size that makes them an outlier for that year. That's about 9 all I can say about that particular case, and it's the only 10 one that I'm aware of. 11 MR. CARROLL: How about 1988? 12 MR. MARION: I don't know the specifics about I. 13 1988, other than looking at the charts that Warren had which i 14 were from material'that we presented at a meeting that we 15 had with the staff. i 16 I believe there were a couple plants identified on 17 that, but they are diffetant plants than the one in 1989. 18 The point is you can look at the data for any given year and 19 you will likely see an outlier, but, from year to year, it 20 won't be the same utility, won't be the same plant. J j 21 That is the point that INPO has continually made 22 to us and it's the point we made to the staff. And because 23-of the statistical variations of the process and data 24 collection, etcetera, our point to the staff is you [O 25 shouldn't focus on the short number of demands a utility has

') 1 76 I b 1 to address in the course of a year; you should look at long-l 2 term performance. 3 MR. CARROLL: Or a situation where somebody's J 4 performance deteriorates, have they responded to what caused 5 the deterioration. 6 MR. MARION: Exactly. The second element that we l i 7 believe provides resolution of the generic issue is a 8 modification that we had made to one of the station blackout 9 initiatives. It was Initiative 5-A. As we modified it, we s 10 identified it as -- I'm sorry -- Initiative 5. As we 11 modified it, we identified it as 5-A and we basically added 12 ths trigger values that were identified in Warren's 13 presentation. ( 14 Let me just take a minute and indicate that our l 15 Appendix D document relates to remedial actions to be taken 16 when you exceed these triggers of 20, 50 and 100 demands. 17 Our program is based upon the 20 demand as being a 18 performance indicator, not a reliability indicator, !9 basically because of the small demand size. 20 Our reliability indicators are based on the P 21 combination of the 50 and 100 demand sample, and this is the 22 double-trigger exceedance. If you exceed thece two values, t 23 you get into some heavy comparisons to your programmatic [, - activities that are used to support your day-to-day 24 l 25 maintenance, surveillance, etcetera, and I will touch on L

77 1 that a little bit later. ') 2-But in March, the NUMARC Board of Directors j 3 approved this initiative and this initiative was put in 4 place to basically direct a focus on diesel generator 5 performance as opposed to programs. We felt that a 6 performance-oriented approach was the way to go and because l 7 of the fact that we were able to reach a consensus with the 8 statisticians on both sides, those that Al Sarkiz referred 9 to that the NRC had contracted, as well as the people that 10 we had used with EPRI, and that these were reasonable-11 consonaus trigger values. I 12 One of the reasons we focused on this is because ( } 13 this was a way that we agreed would be a good scheme to 14 monitor your performance; look at the numbers of failures. i 15 and successes you have over those numbers of demands. That s 16 way, you don't have to worry about the statistical 17 variations of the data. You don't have to worry about the 18 interpretations of the statistics, and you don't have to 19 worry about making point estimates and judging performance-20 as a result of a particular failure that may happen on a 21 diesel generator next week. 22 Our program is strictly based upon focusing on the 23 failures and taking appropriate remedial actions, and that's 24 basically the layout of Appendix D. Secondly, in this jO 25 initiative, it calls for the reduction in accelerated

78 1 testing, which we, quite frankly, feel is a tremendous ) 2 benefit to the industry and we feel by undertaking this 3 reduction that performance of diesel generators in terms of 4 reliability will be sustained. 5 We have, through this initiative, asked utilities 6 to pursue this reduction by implementing changes to the-7 technical specifications. 8 MR. CARROLL: I'm not quite following the four~in 9 25 versus -- that's for individual ones. 10 MR. MARION: That's an individual diesel l 11 generator. Let me indicate that the trigger values are 12 based upon a plant unit because the concerns of reliability l l 13 on station blackout were focused on unit performance. 14 MR. CARROLL: Right. 15 MR. MARION: This four out of 25 is what the staff 16 refers to as a problem diesel generator. Now, these trigger L L 17 values, as you realized this morning through Warren's 18 presentation, have been accepted by the staff and we feel L L -19 that those trigger values, with the process of taking l 20 appropriate remedial actions when you hit a trigger or 21 exceed a trigger, is what is necest ey to maintain and 22 monitor your performance. 23 That, as Bill indicated, constitutes our 24 interpretation of the program or the process, if you will. .jO 25 The guidance in our Appendix D document is basically 1

) l i ) 79 l 1 reflected by this diagram that you saw earlier today, and I b 2 would like to take a moment and walk through this briefly. 3 In that situation, you do what the people are 4 currently doing today, which is respond to the failure, 5 identify the root cause, and take appropriate corrective 6 action. In the case where you have one trigger exceeded, 7 you continue with the root cause analysis, address other' 8 applicable failures tha*. you may have had for that machine 9 or for the same root Jause. 10 And we put a scope of this effort of looking at j 11 the last 100 demands to see if you've had a similar problem i 12 occurring in the past. We feel that looking at that 100 13 demand gives you a better feeling of your long-term 14 performance over that period of time, and 100 demands is 15 typically about four years, three-and-a-half to four years. 16 Additionally, look at your maintenance history in 17 terms of those diesel generators and how you responded to l 18 the problems that you may have discovered previously over l L 19 those 100 demands, identify a failure pattern; a repetition, 20 if you will, a trend, and take appropriate corrective 21 action. 22 MR. CARROLL: Those are all the things Vogtle did 23 on their 70 temperature switch failure. b[ ) 24 MR. MARION: I am not familiar with what Vogtle 25 did or did not do on those sensor failures. lt

80 i 1 MR. RAISEN: Maybe, maybe not, bt,t they were not i V 2 committed to it when that happened. They are now committed 3 to it. 4 MR. MARION: Yes. This commitment is effective 5 March. On the double-trigger exceedance, and this is the 6 combination of the 50 and 100 demand triggers, which were 7 the similar activities that were done on the single trigger, 8 except there's a difference. Program changes. 9 Our point in this is you've had a series of 10 failures, you've' looked at performance over the last 100 11 demands, how do you stack up. You have not been able to 12 control your downward trend of performance on the 50 f f 13 trigger, for example, or it could have been this, and you 14 clearly have hit this one, so things are still going down, 15 they have yet to bottom out, what are you doing. 16-We provide discussion on recommended action in 17 getting an independent assessment of your programmatic 18 activities that support your diesel generators. It's here 19-where we make the link to the Appendix D detail. l. 20 The Appendix D document essentially provides l 21 appropriate guidance for addressing that particular item. I l 22 just want to point out that the Appendix D document was L 23 issued to the NRC in May and Warren touched on that. 24 The information in the Topical Report, which was {O 25' the same information that was in the previous version of the

81 1 -1 Appendix D document, was removed for the reasons that Bill (, s (, 2 had indicated earlier. Let me just summarize those briefly. 3 We felt that because of the established industry \\ 4 performance, and this is the performance that the NRC has ) 5 accepted, that we felt that that level of detail was too 6 prescriptive. The obvious question that came to us is why ^ 7 do we have to do this. 8 We had difficulty answering that question when a ) 9 plant has acceptable diesel generator reliability. We felt 10 that that change _in focus of that information in the Topical 11 Report was necessary because we believe that the staff was 12 in the process of focusing on performance rather than ) 13 prescriptive programs. f l ) 14 Let's look at the indicators. Let's not worry l 15 about how you do that. Let's look at the indicators. If 16 you're meeting the indicators, then the how question doesn't 17 come into play. 18 MR. LEWIS: Could I just ask one question for 19 clarification? I'm a little confused in terms of what you 20 look for. On the trigger, say, three out of 20 for 95 l-21 percent, is that three out of any 20; that is, does the l 22 counting begin when there is a failure, so it's really a 23 test for whether there are two in the succeeding 19 sach 24 time there's a failure? gO 25 MR. MARION: Yes. L

82 1 MR. LEWIS: So it's really not three out of 20; 2 it's two out of 19 and the counting starts whenever you have 3 a failure. 4 MR. MARION: Yes. 5 MR. LEWIS: Now, the a priori probability of that 6 is rather high, even with 95 percent reliability. Thank l 7 you. That's what I needed to know. 8 MR. MARION: And that's why we felt that in terms 9 of these triggers, we had to come up with some mechanism 10 that we can use so everyone would be able to gauge the 11 effectiveness of meeting those targets. We came up with 12 those triggers basically because the statisticians agreed. I 13 MR. LEWIS: Well, that remains to be seen. I I 14 cannot believe that. 15 MR. CARROLL: To what degree of confidence? 16 MR. LEWIS: If you have 95 percent probability, 17 every time you'have a failure, the probability that you'll 18 have two more in the next 19 is around 20 parcent. So it's 19 going to happen over and over-and over again and it won't 20 mean a thing. 21 MR. MARION: That's why we identified the single 22 trigger exceedance or the failures in 20. 23 MR. LEWIS: I didn't mean to make you go 24 backwards. ( 25 MR. MARION: That's fine. That's why we identify

) 83 1 this clearly in our document as a performance indicator as 2 opposed to reliability, because of the demand sample size 3 and the fact that we didn't want to get people knee-jerking 4 into situations of restructuring their programs because of a i i 5 couple failures on a machine. 6 MR. LEWIS: I have to shut up. Otherwise, my 7 Subcommittee Chairman is going to shoot me. 8 MR. MARION: With regard to the Topical Report, 9 the third item, as Bill indicated, are reasons for 10 separating that from the appendix and not submitting it to 11 the NRC, was the fact that the staff would use it to support 12 inspections independent of diesel performance. i 13 We feel that that's an inappropriate use of 14 industry as well as NRC resources. 15 MR. CARROLL: I would have probably added a fourth 16 bullet. I thought the notion that if you didn't do it 17 exactly as prescribed, you had to do an analysis to justify ~ 18 why you're doing it differently, I would find very onerous. 19 MR. MARION: I agree with that. From the 20 standpoint of getting guidance out to the industry, NUMARC '21 held a workshop in May of this year on the initiative, the 22 Appendix D document, and the Topical Report. 23 Those documents were made available to the l 3 industry as a result of the initiative process in May. The 24 1' 25 documents went out to utilities. I'm sorry. In April. And 'e-V

84 -1 in May at the workshop, the documents were given to the (r~)- i ( 2 attendees, so that guidance is out there and the commitment i' 3 to use these triggers is in place through the NUMARC 4 initiative process. 5 As Bill indicated earlier, we had provided copies I 6-of our comments on the staff resolution package for the B-56 7 issue. I just want to briefly summarize our reasons, some I 8 of the major reasons for opposition of the staff's proposed j 9 resolution. 1 10 We felt that the prescriptive guidance was l 11 unnecessary and unwarranted in light of the acknowledged ] i 12 industry performance in terms of diesel reliability. As ~N 13 Bill indicated, we had concerns about invoking the 50.54 (f) '14 letter or the 50.54(f) process to extract a commitment. t L" 15 We felt that this bypassed the regulatory process i 16 and, indeed, established guidance positions in the 17 -Regulatory Guide as regulatory requirements. We felt also 18 in the regulatory cap -- I'm sorry -- in the resolution 19-package, there was a discussion that linked the approval of 20 the tech spec submittal on the reduction in accelerated .21 testing to a commitment to the Regulatory Guide positions, 22 and we felt that that was totally inappropriate. 23 The staff had also indicated in their package that 1 q they had performed a backfit analysis and they relied on 24 25 what was provided to support the Station Blackout Rule. We i e- w ,-w-m-- w,* pn - - - w w-r as_,._,,____

85 1 reviewed that NUREG document and concluded that the staff 2 did not quantify the impact of diesel reliability programs 3 and we felt that because that was not quantified in that 4 NUREG document that the 50.109 backfit rule requirements'had 5 not been met. 6-We also felt that the package had a clear direct 7 focus on programs independent of performance. I'd like to 8 take a minute to just summarize the actions that industry 9 has undertaken to basically provide the NRC with everything 10 that we believe they need to close the B-56 issue. 11 The utilities have provided a docketed commitment-12 to maintain the target reliability. This was discussed (( ) 13 earlier by Bill. Have a commitment to use the accepted set 14 of trigger values to monitor the target reliability, and 15 this is captured in our initiative, and the trigger values 16 are also captured and agreed to by the NRC, so that's in 17 place. 18 We believe that through our Appendix D document, 19 the NRC will be provided information on individual failures 20 and associated corrective actions. We feel that they'll 21 also be provided information on combinations of these 22 individual failurer,as they relate to the plant unit 23 performance and.eliability. 24 Additionally, in the case where a utility hits the 25 double exceedance trigger situation, the NRC will be l l v.

e h 86 -l' provided information on the comprehensive programmatic 2 improvements that are being undertaken by_the utility in 3 that particular case. 4 Therefore, we believe that these actions are +: 5 sufficient and-are all that is necessary for closure of the ~ 6 B-56 Assue. The intended reliability goal of.95 has been 7 met, satisfied, and' superceded. O With that, I have finished my comments. Thank h.", -= 9 'you. 10 MR. MINNERS: May I ask a question for p . 11 clarification on the last slide? 12 MR. WYLIE: Sura. Ij 113 MR. MARION: On the last slide? 14 MR. MINNERS: Yes, the last slide. I don't 15 ' understand your last three bullets that we're going to get ,v 16 tnat information. You're going v colunteer to give that 17-information? Because it's-certainly not required, pardon 18 the expression, by the Reg Guide. 19 MR. MARION: That is the information that is at 20 the plants. 21 MP.. MIENERS: Okay. Fine. So that would be on-i ?. site infcrmation. Fine. k 23 MR. WYLIE: Let me see if I understand, Alex. The 1 24 reason that you chose to take what was in Appendix D and put 25 in a Topical Report is that you feel that the individual -- ,g ,a [

87 lL ) this basically is advice to the utility and the utilities k ) .2' may have-different programs that accomplish the sams and 3 results. 4 MR. MARION: The appendix -- that's correct. We 5 felt'that that level of detail should only come into play 6 when the utility has established poor performance in terms 7 of diesel reliability. 8 At the point that 50 and 100 demand triggers are 9 exceeded, that's an appropriate time to ask the questions 10 about how they've been conducting the root cause analyses, t 11 how they've been conducting their maintenance and 12 surveillance activities, how are they conducting their } 13 - activities, their corrective actions, etcetera. 14 And our Topical Report provides some suggestive 15 considerations for the utility as information when you get 1 16 to'that point. In our workshops, we indicated to the-17 utilities that this was just information that was gleaned 11 8 from several people who put it together'and there are other s 19 ways, other quantitative techniques to assess your 20' performance and respond to a declining trend. 21 Because of its informational content, we felt more 22 comfortable with removing that in total from Appendix D and 23 ' publishing it as a Topical. '24. MR. WYLIE: There are certain elements of what's .. ( 25 proposed in Appendix D that probably wouldn't change. a L ww.

<v -88 IL They're just good approaches to doing certain things. But

g; f.i,[fs,)$

2 there'are certain areas, it would appear to me, that may y 3 -change with' time. '4 For example, I didn't find much written in there,. 5~ particularly in the maintenance area, that spoke to 6-preventive maintenance testing, such as -- I'll just use one 17 of the examples -- such as installation power factor testing i 8-on generators. 9 In fact, there is nothing in there in the '10 ' electrical area. If you're going to write something that 11 says this is the way you're going to do it, you've got to be ~ 12 all inclusive'and it's-going to take a much larger, more 9 ) Vast document than you've got here. 13 14 So I don't see how you can adopt something that is 15 so -- because each utility has got preventive maintenance 16 testing programs that are quite extensive and they're not l 17-documented, a lot of them. L 18 MR. MARION: Our process calls for utilities to 19 only reassess those actions, those programs, PM activities, t 20 if those PM activities were put in place in response to 21 particular problems that resulted in failures-of the diesel-L L 22 generators. '23 And unless you have those failures and you don't 24. have anything that comes to surface that questions those. l 25 activities, we're not asking utilities to look at them under 1 s i

89 y 1: - this particular program. ' O 2 Now, we don't know how this maintenance issue is 3 going to evolve finally and what impact it's going to-have 4- - on some' aspects of this. But you bring up a good point, 5 because'one of the things that we found here wr.s that each 6 of'the diesel generator manufacturers 1..: different ,7; requirements on their machines, and the requirements of 8 thmu machines in terms of maintenance, etcetera, are-9 adjusted with specific experience and performance of the - 10 utilities. 11 We find out that there is such a variability in l-12-terms of that level of detail.and the reasons they conduct 13-certain actions in terms of maintenance, it was very 14 difficult to capture that as one set of generic tifings to - 15 do. -We had a great difficulty.with coming up with'a 16 consensus in that regard. 17-MR. WYLIE: Do you have a comment? l-18 MR. CARROLL: I'm still troubled by Warren's Slide ^ 19 19 on Vogtle, which looks to me like a fair summary of what 20' I read in the IIT report. Would the situation at Vogtle, if il 21 they had in place this program, would that have been caught? [ 22 That seemed to me to be a beautiful example of some very l 23 poor maintenance. b 24 MR. ROSA: I think that can be explained by the - h,j 1^ L . 25 fact that these temperature sensors failures that were m f d g/ g 9 m =

L i ..e s 90' /' 1-experienced did not result in the diesel generator --- l 'I 2' MR._ CARROLL: They did at one time. 13 MR. ROSA:' Those triggers would not have been-4 -activated. 5 MR.' CARROLL: Until-the actual failure. 6 MR. ROSA: That's'right. 7 MR.'MARION: Right. 8' MR.-ROSA: Even then, that didn't trigger -- -9 -MR.'MARION: Right. And it's my understanding j 10 that the Vogtle reliability was high,.9899 at that time, 11 based upon the INPO plant performance indicator program data g 12 -collection process. 13 MR. ROSA: That's right. 14 MR. CARROLL: So what does all of this-say about-4 i-15 the afficacyLof the program that everybody is saying is -- 15 MR. ROSA: The monitoring program or the L17 reliability -- 18 MR. CARROLL: Both pieces'of it. i 19 MR. MARION: The' maintenance program? Are we here

2 0 '

to discuss maintenance programs?

21 '

MR. MINNERS: That's part of the-reliability 22-program.- That's the' question. I don't know how you answer i 23-that, Mr. Carroll, of whether if they had a difference }() 24 maintenance program or we outlined for a liability program, 25 if they would have done better. I don't know how you

1 -t 91' -1: demonstrate that, tp $_,/. 2' Excuse me for interrupting your presentation, but-3 .I would like to make one comment. People are characterizing 4 position C-6 as detailed prescriptive requirements. I just -5 don't see them that way. I mean, I-think they're very broad 6 generalizations. 7 If you look at the most odious'one, which is'EDG-8 Maintenance Program,-just to pick one that overybody cantget r 9 hot under the collar about, it tells you to do vendor-10 recommended maintenance actions. That's not very ' ll prescriptive. 12 MR. CARROLL: But I've seen vendor recommended L( . 13 maintenance-actions that are no good. I want to do j 14 something different, l f -15 MR.'MINNERS: Then you ought to have.a good reason Du 16 for doing it different. i 17 MR. CARROLL: Yes. I probably do. i i: 1 18 MR. MINNERS: You shouldn't just say I'm not going 19' toLdo it. :You ought to have' someplace where somebody has' l [ 20 sat down and said, hey, that's a' dumb thing to do, I~ain't i 1 I '21 going to do it. That's all it says. It should be based on 22 that. It doesn't say you have to do it that way. 23 It says you ought to prioritize your maintenance . 24 actions. I don't think that's prescriptive. That's the ~ 25 general rule. It doesn't say how to prioritize it. + l-. 1

i 92 s (1-MR. CARROLL: No, but a resident inspector could li"'k 3_,/ 2 waste an awful lot.of a plant maintenance organization's. 4 3 time saying show me how you're doing this. 4 MR. MINNERS: I've had this dis ,sion with other 5' people and I feel a little put upon that the part of the 6 staff that is writing the guidance, people keep coming back t 7 and saying, well, that's going to be misused by the 8 inspection staff. 9 That shouldn't be our problem. -If we have a 10 problem with the inspection staff, but I'm not sure we do, 'll that ought to be fixed there and not by telling people, hey, 12 don't~ write stuff. ( } 13 MR. KERR: This bulkanization of the NRC into 14 separate little featureo where nobody's one unit is 15-responsible for what is done in the other unit is what 16 .causes a lot of difficulty, Warren. I know it's easier not 9 17 to have to worry about how it's going to be' interpreted, but 18' noi'. it's going to be interpreted will finally determine 19 'whether it's any good or not to the operating plant. I L 20 If you don't have some idea of how it's going to 21 be interpreted, how can you know whether it will accomplish 22 your objectives. 23 MR. MINNERS: Well, I know how it should be 24 interpreted. I don't think it's reasonable to expect me to 25 go beyond that point. And if it's being interpreted

~ 93 -1 incorrectly, then people ought to go to that-section of the -2 'NRC and say shape up your act; don't drag me upyin front of 3 here and. pummel me because'the inspectors are doing wrong. 4 And I'm not-sure they are, but I can't do anything -5 about it. You may rot like the bulkanization, but it's an 6 organization and I can't be responsible for.everything in 7: the organization. If you have a problem with another part m 8 of the organization,' bring those people down here and give SF them the business. 10 MR. KERR: And they will say, well, those guys =11 that wrote this regulation wrote it and that's the way it's 12 supposed to be interpreted, l( ) 13 KR. MINNERS:-' If that's the case, fine, but that 14 isn't what people are saying. I don't think you can point 15 out in this regulation that a guy should be doing that or 16 that we should be giving better guidance to our inspectors 17 -which say, hey, you know, don't do it this way, here's how 18 you ought to do it. 19 MR. KERR: It's a tough problem. 1. 20 MR. MINNERS: But the other side of the coin, Dr. 21 Kerr, is if you don't give the inspectors any guidance, if 22 you j st give them very vague generalities, then what does 23 the inspector do? I think that leaves the inspector up to j 24 his own devices to do whatever he wants. 25 If nobody tells him and gives him direction in 1

i l t 94 1) specifics of what he's supposed to do, what-is he supposed. f-sg J-) l. 2-to'do? Ha's going-to take his own ideas and do-it,.and can 3 you blame him for that? I think it's better to give him -4~ ~ specific guidance and say, hey, this is what the agency ] 5 thinks you ought to do. 6 MR. KERR: We weren't arguing that inspectors 4 7 shouldn't be given specific guidance. I wasn't.- I was i 8 saying -- l 9 MR. MINNER:1: If you take specificity out of Reg 10 Guides, you give them less guidance. 7 11 MR. KERR: Wait a minute. You can take 12 specificity ~dt of a Reg Guide by not even mentioning 13 maintenance and simply hav ing performance specifications. l 14 MR. MINNERS: That's a different argument. 15 MR. KERR: I agree. u 16 MR. MINNERS: That's a different argument, and 17 that's a correct argument. Is-this necessary,odo we need + J 18 something beyond the performance monitoring? I agree that's 19 a different argument. But to say that we shouldn't have 20 this thing in here because it's going to be misused, I think .21 that that's not a good reason for getting rid of it. g 22'- And I don't doubt that the NRC doos things. I've 23 done things in the past to the industry I wish I hadn't i 'i 4 done. I think they were stupid. But that doesn't mean that ) 25 .everything we do should be thrown out because people aren't + ' \\i fk

'95- =1 perfect. 2 MR. WYLIE: Let me ask a question. Warren, in-3: taking Section C-6 and in putting it in an appendix, is it p 4 going to be characterized as the staff considers valid 5 reliability program to contain these principal elements or 6 how is it going to be characterized? 7-MR. MINNERS: I don't know. The handout hau une l 8; words wnich says it's illustrative and examples of what you 9 can have. -10 MR..WYLIE:. I mean, this is just an outline, 11 right? What you handed out is just an outline of what the. i 12 appendix is. xi 13 MR. MINNERS: No. This is the proposed revision ,(() 11 4 of Section C -- 15 MR. WYLIE:' -You're not going to take all of these 16-words-that are now in the Reg Guide and put them out here? l 1 17 -MR.' MINNERS: We gave you a revision to Section 6. l 18 MR. WYLIE: Yes. 19 701. MINNERS: It's 6.2 through 6.7. We're going 20 to take those practically verbatim and put them in the 21 appendix. 22 MR. WYLIE: That's what I'm asking. You're going 23 to take all of these words that are over here -- and they 24 refer to tables, like Table 5, for example. This says i .O 25 typical example of preventive maintenance activities. l

96 l: ^ 1: MR. MINNERS: Yes. 12 MR. WYLIE: Which is not all inclusive. 3 MR. MINNERS: That's correct. It says typical-4- activities. 5-MR. WYLIE:- Okay. That's what I understand. This-6: is-just an outline of what you are going to do, and you're ' J7 going to take the words out of the body of the document. x 8 MR. MINNERS: C-6 is exactly what'we're going to 9 do,.how we're going to review section C-6. This is what we 10 proposed. / 11 MR. WYLIE: Maybe I misunderstood. I thought you i I -12 said that what you were going to do is, for example, take ({ 13 surveillance plan, that you were going to take 6.2 as 14 written-and put it in.the appendix. H15 MR. MINNERS: That's correct. 16 MR. WYLIE: Okay. -l 1 17 MR. MINNERS: This piece of paper -- .18. MR. WYLIE:- This is an outline, then, of what 19 you're going to do. 20-MR. MINNERS: No. This is exactly what we're 21 going to rewrite. C-6 is still going to be in there, Mr. 22 Wylie. 23 MR. ROSA: The statement is on the backside, Dr. i 24 Wylie. It says Appendix A provides illustrative examples in g 25 consideration. That Appendix A that's on the backside of

97 1 your sheet contains the illustrative examples. I 2 MR. WYLIE: See, C-6 would be just this. 3 MR. MINNERS: And everything that's in there now 4 sill go into the appendix. 5 MR. WYLIE: Okay. 6 MR. MARION: The concern that we have is that an 7 inspector would come in and look at that set of surveillance 8 activities and ask the utility to justify why they are not 9 doing each and every one. 10 MR. MINNERS: Now, should he be doing that or 11 should he not be doing that? 12 MR. WYLIE: Should he be doing more? ( 13 MR. MINNERS: That's possible, too. 14 MR. WYLIE: I mentioned there is nothing in here 15 really on electrical maintenance. 16 MR. SERKIZ: That's correct. Neither did the 17 industry identify, in the course of our interactions, that 18 was important. 19 MR. WYLIE: But it is, they do it. They do it. 20 MR. MINNERS: In the sense that he's supposed to 21 do the vendor recommended maintenance actions and if there 22 are electrical-related maintenance actions, l 23 MR. WYLIE: I mean, if I wanted to know whether a l 24 diesel generator was going to fail tomorrow, I would g 25 certainly periodically look at the installation power factor

z 6 98-measurements,and trend those. They do that. The better 1- - (I^) L_/ 2-ones I-know do. j ~ ~3; MR. SERKIZ: We provided for the Committee some 4-illustrations'of what we call maintenance issues andean: t 5- . inspection team has gone out. We've identified for you some 6' of'the recent inspections. Wo.have a plant in here which 7 was givec a Level 3 violation. 7 8 In a Regulatory Guide, we are trying to lay'down a 9 common set of guidance. The staff is trying to provide,.in 10 a singular place, this guidance. The industry or NUMARC has 11 stated that they have commitments out there and so on. / [ 12 I guess the question I have and it's one that'I've N;' 13 repeatedly raised with NUMARC,-is-the NUMARC initiative and [ Lc 14 the guidance that.goes with it a licensing commitment, and r + .\\ 15 the answer I've gotten consistently is no. g 16 The. 50. 54 ( f) letter is'a mechanism that says are f "17-you going to do this for the NUMARC Initiative.5-A and their H 18 Appendix D, which corresponds.to our position C-3,.41and 5 - L i l' 19 excuse me -- C-5 we backed off:even, and said give us a J

)

a 20 special1 report on the problem diesel. 2 11 It provides a position C-6 in diminutive form of 1 22 guidance that's out there in a much more Extensive Topical 23 Report which is going to be used for guidalce, training or 24-what all, and I think the issue on the table is do you g,-sg, 9 25 provide it in a Regulatory Guide and ask fcr a licensee to L - -- a

99 L1 tell us that's what he's going to do or tell us what he's-21 going to do instead. 3 MR. LEWIS: I don't understand the last sentence, j u 4 because although the' paragraph says that, in fact, the' 5 Regulatory Guide has got it and you could do it in.some ] 6 other :way, that's okay. It then goes on to say whatever ~ 7 that other way is, it should be reviewed for consistency 8 with this one. 9 MR, SERKIZ: Which piece of paper are you on? .10 MR. CARROLL: Last paragraph, Page.1. 11 MR. SERKIZ: Okay. y 12 MR.-LEWIS: If you're going to offer other l L13 options, why should you have to review them to make sure i' 14 they're consistent with this option? 15 MR. SERKIZ: It's consistency with Regulatory 115 n, 16 .and this Regulatory Guide. I 17 MR. MINNERS: We're saying we've given examples, 18 details, you could use those.if you want to. If you don't '19 use those de c.ils,, provide your own and then compare it to s I 20 the Regulatory Guide which has only very broad criteria. 21. That's all we're asking to do. L '22 MR. LEWIS: It says review for consistency. Does 23 that mean you better make it consistent, in which case it's 24 not an alternate -- q 25 MR. MINNERS: I guess the presumption was that the ) i

100- ~' -(..

1 utility has cosnitted to the Regulatt,ry Guide, which are the

) A/ L2. positions in Section C.- 'But he says I'm not going to follow 3 Appendix A. He,said that's okay, that's allowed. Show us '4 how,.what you did. 5 MR. LEWIS: Is it clear that those options you've ^ 6

just described will-be clearer to the licenses than they are 7

.to me?- 8-MR.-MINNERS: No. I think it.makes.it very vague. 9 You haven't told me the specifics, and that's what people '20 are complaining about. We're in the same problem. If we

11-give people' general guidance,'they say, gee, tell me what-12 you want.

If we give them specific guidance, they say, oh, L i( '13 that's too specific. We've always got that problem. That's t l. ' '14 what this problem is here. 151 We can either tell them exactly what we want and L 16 they'have no: flexibility, or we can give them flexibility -[ _17 .and there has to be uncertainty as'to whether they'll accept L 18 it. 19 MR. LEWIS: You're crying to the wrong people. 20 You have to be explicit about what you say, because you're a p ~ 21 regulatory agency. c 22 MR. MINNERS: We're being tnld by the industry L 23 that we shouldn't be explicit. ( ~2 4 MR. KERR: You're being told by the industry that n 25 you're trying to regulate things that you shouldn't try to r x-

101-l l' . regulate.. That's what you're being told by the industry. hNL 4,1 2 MR. WYLIE: Okay. .Dr. Kerr had a question.

3 MR. KERR

In a memo from Mr. Todanti to somebody, 7 1 4 I've. forgotten who, apparently there's a concern, Mr.. 5-Minners, about the reporting requirement, and I wanted to 6 see if I could understand what is different about what is t 7 wanted than what NUMARC was proposing. 8' Is this. simply an earlier trigger than NUMARC is 9 proposing or does it have anything to do with what NUMARC is 10 proposing?- It's the July 6 memo from Mr. Todanti to;-- 11 MR. MINNERS: I remember that. 12 MR. KERR: It wasn't clear to me what the problem -I 13 was. V. 14~ MR. MINNERS: I'll tell you what the problem was.- 15 The problem was that I crossed out any reporting r'quirement' e 16 and NRR objected to that. They said no, we want to have this minimum reporting requirement. That's what Dr. Todanti 17. 18 is complaining about. .19 : It's that when I revised it and sent-the thing 1, L 20 down to you people, there was ru) reporting requirement at 21 all in the Reg Guide, and NRR said, wait a minute, we've got L. -22 to have something. l 23 MR. KERR: So it wasn't that NUMARC was trying to 24. avoid -- they also had some reporting requirements. q _s 25 MR. MINNERS: I don't know. I don't think they 1 l

102 .~ 1 ' objected'to that reporting requirement. I'm not sure. .(N) _( 2 MR. -. KERR: All right. 3 .MR. SERKIZ: On Page 20 of the Regulatory Guide 4 that was provided in the initial B-56 package, it said when 5 reporting EDG fail'ures, all provisions of'10'CFR 50.72, .6 which is the immediate notifications, 50.73 which is LERs, 7 and 10 CFR 21 which is just defect and noncompliance, plant 8 technical specification and other currently NRC reporting-- 9 regulations. 10 ' Sort of a minimum wording, go by.the standard 10 i 11: CFR or whatever. Now, Todanti said no, we had agreed before I ! that we want-to know when you get a problem diesel. ( 13. MR. KERR: I understand. 4 14 MR. SERKIZ: The industry, in their Appendix D,. 15 has very similar language on reporting the underlying causes 16 that we show for the problem diesel. We say that we are 17 less prescriptive or requiring less than the industry shows-1 18-in their Appendix D because the staff agreed that if we are n 19 going to have a special report that comes into the NRC, 20 we'll dwell with the problem diesel. hl' MR. KERR: Thank you. T 22 MR..WYLIE: All right. Has the Subcommittee any 23 recommendations as to what we tell the Full Committee? I 24 would-suggest that we have the staff and NUMARC come in and 'i p' ~ 25 make presentations to the Full Committee tomorrow. We don't 4

103 l' have'quite as much time. We've got one hour and 45 minutes, O k-[ ,2 so we need to condense it somewhat, if we can maybe f.ot get 1 3 into some of the -- 14-MR. CARROLL: I'd like both of those entities to 5 think a little bit more about the Vogtle situation and 6 whether a future situation having those same elements is 7 going to be picked up by this programLthat industry and NRC 8 are supporting. 9 That's really the first major failure of a diesel i 10 generator that I.can think of when it was desperately 11 needed. I mean, we've had failures, but they haven't been. 12 during a time of need. You wanted to say something? (I ) 13 MR. ROSA: I want to request permission to express i 14 my-minority opinion'with regard to one specific point, and

15 that's the definition of a problem diesel.

-It's the view in i 16 my branch that three out of 20, not four out of 25, should 17 define.a problem' diesel.- 18 I understand the statistical 1 uncertainties that 19 have been discussed here, but when you have a problem 20 diesel, th:.re out of 20 seems reasonable. Those three out 21 of 20 could be three out of the last five, in which case it l 22 . would.not be a problem diesel under.the present definition. l l L 23 The only thing that this would do if you changed 24-that from four out of 25 to three out of 20 would simply 25 initiate the accelerated seven weekly tests and that report. 1 t m

q: 104 QL' The NRR, as an: office, fully supports the presentation as 12' presented by Warren. The opinion I've expressed within my q s; 3 branch by myself. personally. of 4 L MR. WYLIE: Thank you. P c =5 MR. LEWIS: Let me just say something about1three 4 6 cut of 20. If the underlying probability of success is 95 ( 7 percent, and three~out of 20, the counting starts every time 8 there is a! failure, then there's a 25 percent probability on ~9 statistics alone that at any time there's-a failure, it will 10-be followed'by two more in the next 19; 25 percent. 11 1. So for a' plant that is doing a lot of testing-, 12 this is going to happen once a year without any implication j<13' that there is a real underlying problem. That's just raw U L 114-statistics. 15' MR. WYLIE: All right. Does that give the staff 16-enough guidance for tomorrow? } L '17 but. MINNERS: So we're talking about a 35 minute 18 presentation? L19 MR. CARROLL: I think we ought to tell them to get L 20 together with NUMARC and reach an agreement with NUMARC l.: ' 21 tomorrow. lt 22 MR. SERKIZ: If it hasn't worked in the last 26 L e 23 months -- ^ 24 MR. CARROLL: But you've got a lot more background 25 now than you did 26 months ago. t 4 i g <n

...-~ -r 105 -l 1> MR. SERKIZ: No more since last November than I do i Q,s,- f 2 now. -- i 3- -)GR. KERR: Yes, you do.. You had the Vogtle 4-failure. 5 '- MR. SERKIZ: Well, that's why we provided you 6 gentlemen some other examples. ' 7. MR. LEWIS: The first sentence-says the principal .' 8. elements of a reliable program are composed of the.following j 1 9 principal elements. I would fix that, t 10 MR. SERKIZ: We'll take care of the typographical. 111 MR. WYLIE: Let's adjourn this meeting and we will 12 hold the next meeting at 1:30. I: ) -13 '[Whereupon, at 12:26 p.m., the meeting was r0 ~14 adjourned.]. 15 16 17 18 19 20 t 21 22 u 23 [( s 24 25 I' =>

, nq,w + [I REPORTER'S Ct'.RTIFICATE This.is to certify that tha attached. proceed-ings before-the United States Nuclear. -Regulatory Commission in-the matter oft n-NAME OF PROCEEDING: AC/DC Power Systems Reliability-Subcommittee DOCKET NUMBER: PLACE 0F PROCEEDING:. Bethe'ada, Maryland. -were held as herein appears, and that this is the original transcript thereof-for the file of the United-States Nuclear Regulatory Commission taken by'me and thereafter reduced co typewriting- ~ by me or under the direction of the court report-ing company. and that the transcript is'a true-and accurate record of.the foregoing proccadings.- f& _e Official-Reporter Ann Riley & Associates, Ltd. 'O

J t g 91 X' -( L'O 1 -RESOLUTION OF GSI B 56-1 PRESENTATION TO THE ACRS SUBCOMMITTEE m> ON AC/DC_ POWER SYSTEMS RELIABILITY 1 August 8,1990- -W. Minners RES/DSIR 4 MS NL/S 360 EXT. 23900 \\ .} '. i v LO; L L o 7

1 T. B-56 RESOLUTION CHRONOLOGY SBO RULEISSUED 6/88 CRGR MTGS NO. 144 & 146-8 & 9/88 RG 1.9, REV. 3: ISSUED FOR COMMENT 11/88-COMMENT PERIOD CLOSED 3/89 MTGS WITH'NUMARC (7 MTGS) 6/89 CRGR MTG NO.164 6/89 MTGS WITH NUMARC (4 MTGS) 7-10/89 CRGR MTG NO 171 10/89 D CRGR MTG NO.176 12/89 ~ ACRS MEETING 2/90 DISCUSSIONS WITH NUMARC 1-3-90 NUMARC SUBMITTAL OF INITIATIVE 5A & NUMARC-8700, APPENDIX D. 5/90 CRGR MEETING 190 7-25,90 ACRS MEETINGS SCHEDULED 8-8-90 8-9-90 O

'6 s+ [o; OVERVIEW RG 1.9, REV. 3 1 1. Consolidates into a single RG guidance previously provided in RG 1.9, Rev. 2, RG 1.108 and GL 84-15, L .thereby minimizing regulatory confusion. 2. Better defines testing requirements, eliminates cold 1 L -fast starts and limits accelerated testing to the - L '" problem" EDG. L 3. Provides common guidance for monitoring EDG reliability levels and actions to be taken. 4. Defines the elements of an EDG reliability program L. and provides illustrative examples of proven _O considerations and practices; suppiaments ouidance provided in RG 1.155, " Station ~ Blackout". 5. Incorporates proven industry practices and is consistent with NUMARC's Appendix D (5-2-90) and related Topical Report. Provides spebific guidance for use by the staff or L 6. industry to review the adequacy of diesel generator reliability programs. consistent with the resolution of USI A-44. u LO slide s l:; o

i 14 -v 4 (7-23-90) TABLE 1 CROSS-REFERENCE'BETWEEN REGULATORY GUIDE 1.9, REV. 3 AND NUMARC-87-00, APPENDIX D (5-2-90) RG 1.9,REV 3 NUMARC-870U SECIION-APPENDIX D Section A, Introduction None (Use RG 1. 9, Rev.3) Section B, Discussion None (Use RG 1.9,Rev.3) Section C, Regulatory Position .1 Design Considerations None (Use RG 1.9,Rev.3) 2 Diesel Generator Testing 2.1 Definitions D.1 2.2 Test. Descriptions None (Use RG 1.9,Rev.3) 2.3.Preoperational and Surveillance Testing None (Use RG 1.9,Rev.3) 3 EDG Reliability Goala and D.2 Monitoring 3.1-Reliability Goals for SBO Introduction 3.2 EDG Reliability Monitoring D.2.2, D.2.3 O-3.3 Maintaining EDG Reliability D.2.1,D.2.3,D.2.4,D.2.5 3.4 Problem EDG ' D.2.4.4 4 Record keeping Guidance D.2.4.6 5 Reporting Criteria-Use RG 1.9, Rev. 3 f 6 EDG Reliability <, Program-Introduction 6.1 Monitoring EDG Reliability D.2 6.2 EDG Surveillance Plan None (Use RG 1.9,Rev.3) 6.3 EDG Performance Monitoring None (Use RG 1. 9, Rev.3) 6.4 EDG Maintenance Program None (Use RG 1.9,Rev.3) 6.5 EDG Failure' Analysis and None (Use RG"1.9,Rev.3) Root Cause Investigation 6.6 EDG Problem Close-out None (Use RG 1.9,Rev.3)- 6.7 EDG Reliability Data System Nc 'se (Use RG 1.9,Rev.3) Section D, Implementation Introduction (Initiative SA) 4 Q Slide 4 1

70 C.6 EDG RELIABILITY PROGRAM

Section C.6 identifies the following principal elements of an EDG reliability program

1. Monitoring nuclear unit EDG reliability levels against SBO targets. 2. Surveillance Plan 3. Performance monitoring of Important parameters. 4. Maintenance Program 5. Failure Analysis g 6. EDG Problem Closeout Process - 7. EDG Reliability Data System These elements are the same as NUMARC's. \\ The RG subsections which follow provide general guidelines (with illustrative examples) for these major program elements. O Siide s

-. k . :1-1. g

o

~ Periodic = W = Testing. Nuclear Unit EDG - Failures / Demands RellatnWtv Trgger Values P 0.95 3/20 5/50 81100 0.975 -3/20 4/50 5/100 New No _ ~ Failure Yes 0 Compare with Trigger Values 1 r 1r 1 r Exceed No Exceed One Exceed 50 and 100 Triggers Trigger Demand Triggers 1r 1 P 1r

1. Root Cause

2. Corrective Actm

2. Applicable Failures

3. Maintenance History

4. Failure Pattern

5. Corrective Acton

5. Program Changes

6. Corrective Acton 1r 1 r os Figure 1 - EDG Reliability Monitoring & Maintenance Activities e

CD

s +. n-2 J FALSE ALARM RATE ' ~(Steady-State Reliability is 98%) ' ' Probability 0.9 - - - - - -= :: - 0.8 0.6 ~~ 0.5 0.4 0.3 - - -W 1 0.2 + 0.1 - - c-2 2 1 0:: 0 25 50 75 100 125 150 Demands (3r20)+(5/50)+(8/100 (5/50)*(8/100) . % m- + (2/20)*(4/50)+(3/20) + (2/20)*(5/50)*(8/100 1 Hig's EDG reesbillly will not result in signifleant false atomes - _4/.). amm.S ~-....

U.: 8 i O O l FALSE ALARM RATE (Steady-State ReliatMty is 95%) Probability 1 em. 09 - (3/20p(5/50k(8/1oo (5/50)-(8/100) ~ ~ i i 0.8 w 0.7 0.6 i 0.5 i i 0.4 1 i i 0.3 i l 0.1 - - - ~ - - - - - - - - - - - - - - - - - - - - - - - s 1 k I

l I

\\ O 25 50 75 100 125 150 1 Demands . oe.g w

Design.ee. "Assp-M i

==g High EDO reIIstMty w111 not resu11 he sigr-mmh l

O O C ~ DETECTION RESPONSE (Rd22%;f Drops from 98% to 92%) Probability 1 m 0.9 ~ 0.8 - - - - - - - - - - I - -- - - -- i t 0.7 ^ ./ -------- OE 1 0.5 0.4 O.3 0.2 0.1 4 I a f f I I Os O 25 50 75 100 125 150 i . o.es nsee. ose ( e o,ogo awso-(3/20)+(5/50)+(W100 (5/50)*(8/100) I t l -b (?>20)*(5/50)*(8/130 -G-(2/20)*(4/50)+(3/20) i l t

1) SV,e Brtipger '.s an In61 celer of onset of degratfellon.

(n_ 2;- onsetton nepo..-- with muerspee ersagers es ssow. -ge l W r -- ~

4 i

o EDG RELIABILITY 1.

EDG reliability situation has improved L 2. Industry reported " Averaged" level is 97 - 98%; the staff has not independently verified this finding. i i 3. Annual plant performance data shows a small number of plants sliding below 95%. O~ 4. Plant specific EDG uaintenance problems . continue to exist.despite reporte1 high levels of EDG reliability (i.e. VOGTLE, COOPER, DC COOK, CALVERT CLIFFS, ZION) s

5. The NRC relles primarily on industry l

reported EDG performance and availability. ~ L I. () slide 10 l l [ +

l 'O i Diese! Generolor Tolol Unreliobility (by slation) Industry Average 0.050- ..w. g70.0W d L emo-uts

14 ut3 ull im im

=> im O Diesel Generolor Tolol Unreliobility (by stolion) T:iree Year Distribution (1/86 - 12/88) n.si Oath W dian ,, 0.004 s.010 o 16-14 $1 $ lotions g 5 io-y l; , = S-4,r t 1t 1 s 0 ~ l. .i. ~ im im .m em im 6. u. w nobnity per statin O Ref 5/2 W89 meeting with NUMARC B 56 Working Group Slide 11

-- _.~.__ l l l m. m. I 19M 1971 1978 E ~g.. t. . x i.. I.. E.. t.. : j.. g..,a p$ e e e e e e ; e ;, c. ... e e e e e. e... d 'I I 1 I ... e. e. . e... m=.. . m.... . m..... O' m. m. 1979 1900 l I.. 2.. I.. e.. t I..A r.. I W ? ,L.. ,L..g I.. 1.. E.. c.. g g n .n l k s,L. A c..... er..... oe.e m a e. vai os.e m. e..v.i m. m. l 1981 1982 .t.. .t.. r.. r.. g g ~ g. g.. i.. r.. L: A rh.. r i.

6..

i .. m..... ... m..... Figure 4.3a Histograms showing emergency diesel generator g failure on demand for 1976 through 1982 NUREG-1032 4-7 Slide 12

t ,l in m m m n a lm 3m e m - m mens l .ei m m ' m m m ' m 'm m',,I,,w' i 3 i PnotatiLffy M FAstunt r tu -,m j sw = in O m m m. m n n lm lm [m F gm e e m m m s m = m w m- ~ m e s i m.a m m m ',w,m m,j,,w ,,,,,,,,,i,i,i,i,r,i,,,i 3 i i i PROSAtlLffY OF FAILURf PA06AtilffY OF FAiwat 1 i Figure 4.3b Histograms showing emergency diesel generator failure on demand for 1983 through 1985 NUREG-1032 48 O. Slide 13 -+w---.s. n- +- -en,- .w v..- ---,-w,

e. w

-w w n., - -,, - -,,,, -

Diesel Generet,r lolo! Uweliability (by station) One Year Distribution (1/86 - 12/86) 0 w ourse M en j ( 26 si s u.a.

is.

s h to-J c s, s 3 i ~l ~ s-i 3 2 s i i 3 1:- d. m 1-e w'4'4'&'4'4'4 se w w .im%,.iw Dieul Generalor 1o1o1 Unrdobility (by slation) One Year Sistribution (1/87 - 12/87) w e m.a. 1 ,em ws N 3 i ao sm n. o 8 ,t0-3-1._34 3 1 1 o I-'- R=Fl,,,,,,, en w a w w ss en w su w sse ..i m w ,.i.n.n Diese10enetolor 10101 Unreliobility (by stolion) One Year Distribution (1/66 - 12/88) w oi,,iu w.6.. 0_M CM1 30

a. _

j 25' s4 sien.n.

20-3

-l15- 'so- + 4 4 3,3 h Y o t.R Sti a 2 $N 82 9M SR LN LN til ..iww .i n v DeM M*f so no m** ting whh NUMAnc a-ca wort.in;. >p

TABLE 6 1 I ALERTS ON EDG'S FROM FIVE PLANTS Plant A Plant B Plant C Plant D Plant E i Number years f '.ure history 6.2 8.1 10.75 4 3.2 Number demands / year / plant 328 36 76 51 366 Number failures in failure history 35 16 18 9 13 Estimated relia-bility per demand .983 .945 .978 .956 / .989 Number of mild O 1 rts e 7 3 3 2 Number of strong

alerts 0

1 2 0 0 s

Strong alerts were received on the EDG reliability histories for Plants B and C.

For Plant B, the EDG reliability appeared to be consistently below ,the target of .95. For Plant C, the EDG reliability appeared to consistently degrade over time, and the strong alerts were received toward the end of the degradation period. These results support the supposition that the alert systems provide alerts on consistently poor performing EDG's,.and on EDG's that are degrading in performance. Ref.: NUREG/CR-4810 5-8 Slide 15

Q Q q 4 Figure 5-7. DEMANDS BETWEEN FAILURES VS FAILURE NUMBER ~ PLANT B Demands Between Failures 1 0 0 0 .......... ~... -.. - -~~ ~~~-- n:nn,n; ;n-*n.:n n*---~ * - - n [ t ... ~ 27l'IIe'I 100 c J m ? y h.z./zi[f l ...........g.................. W4lVD g ... R 10 .- -.... - r- ...... y............. Q..... ..J.. g -- y'l i.@l@. D A. 6............ ). ) hl.t.3 M S I t t I f f I i f I I f 1 .1, 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 u) 4 =g Failure Number l a M = MILD ALERT p% OR 2/20 W 'd8) l e LEGEND: S = STRONG ALERT (2/2e AND SS AND 8/199) Ref.: NUREG/CR-4810 ( 1 h

Figure 5-8. DEMANDS BETWEEN FAILURES ~ VS FAILURE NUMBER PLANT C L Demands Between Failures 1 ooo 100 _......s .. - - -..-. -.............. ~.. -. - -...... ....g. ......................(................g.... ....f 5 I h .....(3 Z \\ N N/lb,) M 4,., ~f, M 10 - - - -- -.\\ ..............--.~.-..-...-.v. Vg.... p ... a. g............... . c. ....9 a. a. t 1 1 I t I I t t 1 I I I I I i 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Failure Number m LEGEND: M = MILD ALERT WJe OR 2/20 AND M O* 'S = STRONG ALERT (2/20 AND SISO AND effee) a" Ref.: NUREG/CR-4810 i i t i 1 , - -. ~ - -

20 0. th PLANT SPECIFIC EDG PERFORMANCE o LERS o Specials Reports Tech Specs RG 1.108, Rev. 2 - Approx 50% of plants need to report 1 o llT (Vogtle llT Report) o Calls to Resident inspectors COOPER NUCLEAR UNIT: ll EDG 1 2 failures in last 25 tests EDG 2 1 failure In last 25 tests Nuclear Unit: ' 4 failures in last 100 tests l0 ca'vear c' ires: EDG 11 98 % EDG 12 99 % EDG 21 99% VOGTLE: EDG-B 5 failures in last 71 valid demands o INSPECTION REPORTS ll L L Q' Silde 18 t i s

n Voatie EDG Maintenance issues 1 .10 Cal-Con temperature sensors malfunction caused DG m o failure & resu'.tod in station black-out. o Cal-Con sensors have been a long standing problem. Approx. 70 sensor failures reporting during i plant history (*). j Only about 5 failures entered into NPRDS. Sensor history was not factored into EDG maintenance program; i.e., there was no m trending associated w/ sensor performance, nor was there any changes to sensor maintenance as a consequence of failures. IIT did not obtain historical data on sensors until 6 weeks into the inspection-material was ( very difficult to retrieve. h (*) All historical data may not have beer. L l retrieved, & there may have been more l than 70 failures of sensors. L. o Vogtle maintenance program for EDGs did not identify temperature sensors as a significant contributor to l potential EDGsfailures - a well organized & functioning l-l program would have done so. l o NPRDS data is inaccurate because of questions L regarding what needs to be reported (see above). .o Failures which occurred during post maintenance-testing were not entered In'o NPRDS, nor viere they e factored in Vogtle Maintenance Program. o -Vogtie was not aggressive regarding EDG problems. (see NUREG 1410) 10 siide is t w .v-.g-..,_.m._ __u-_.w..,.--,,,..-.u -_,___m -w.. w--

] Cooper EDG Maintenance issues Special NRR inspection team noted significant o ' deficiencies in EDG maintenance program. Molntenance program deficiencies noted in lospection Reports 89-11 & 89 12. o inauequate EDG maintenance program lasulted in one EDG being effectively inoperable for an extended period of time. i Licensee was cited for a Level lil violation as a j consequence of the above noted EDG condition. j D.C.- Cook EDG Maintenance issues u 1 Engine overspeed caused failure of a generator & some o damage to diesel engine. O Overspeed caused by improper adjustment of fuel racks o on individual cylinders. . Periodic adjustment of racks to balance engine on the basis of cylinder exhaust temp. No evidence of any investigation.as to why P cylinder temperatures varied from one time period -to another. No evidence of use of prior adjustment data as an Input to current adjustment requirements. No evidence of test & calibration of exhaust temp. l Indicators. l ~ SIlde 20 L u,

( D.C. Cook EDG Maintenance issues (Cont) 3 o Licensee was given poor SALP rating in the area of maintenance due primarily to inadequate. Root Cause determination associated with DG overspeed event. i o Licensee did not Investigate the cause of certain anomalies on parts removed from the diesel engine - anomalies in this case meaning not consistent physical e damage expected from a DG overspeed event. o Corrective actions did not include significant changes to any of the above noted undesirable conditions. Calvert Cliffs EDG Maintenance Program Diagnostic Inspection Team observed maintenance o activities associates with DGs. o Team-inspection did not specifically evaluate the l O-programmatic aspects of DG maintenance, but team did observe problems with the conduct of maintenance activities. Team observations or findings are documented in Inspection Report 89 200. Team member responsible for maintenance observations o believes that past history was not used as an input when determining current maintenance activities, -Poor execution is an indication of serious programmatic o deficiencies. O Slide 21

l ZION EDG MAINTENANCE INSIGHTS q o Observations regarding EDG maintenance (by Region 111 & l Contractor personnel) Trouble shooting of EDG problems S not conducted in i an organized manner using detailed procedures and Instrumentation. ~ ~ it could not be determined how many, if any, of the Zion mechanics assigned to work on EDGs had formal training. No specifics were available regarding what constituted "On the-Job" training & how much each mechanic received. i . Mechanics are assigned to EDG maintenance based on their exposure levels in " hot" areas of the plant, and not on the basis of knowledge; e.g. there is no dedicated EDG maintenance staff. Maintenance procedures are poor, l o Above observed deficiencies were responsible, in part or in total, for 15 failures to start during surveillance testing or trouble shooting in Feb. & March,1990. o Observations are consistent with negative findings regarding other maintenance activities at Zion by a Maintenance inspection Team and a Diagnostic Team. ? UNDERLYING CAUSES s 1) Inadequate root cause analysis. 2) Failure to incorporate operational experience (on site and from other plants) into EDG maintenance activities. 3) Failure to utilize engine parameters for trending purposes. ~4) In-effective problem close out and lack of verification of effectiveness. Q Slide 22

\\.. l 00 STATUS OF RESOLUTIQM l o Staff has followed up on CRGR recommendations. (CRGR Meeting No. 176, 12/20/89) o NUMARC was given the opportunity to submit Appendix D o ' The Staff had discussions with NUMARC (Jan-Mar 1990). o NUMARC submitted initiative 5A and a revised Appendix D (reduced in scope) on 5 3 90. ( o Staff has revised RG 1.9, Rev. 3 to reference NUMARC's Appendix D (5 2 90) as appropriate and Included guidance for an EDG reliability program (C.6) in the RG. A 50.54(f) letter has been prepared to determine the o course of action licensees and applicants plan to pursue and ' suggests submittal of Tech Spec changes to take advantage of relaxations afforded, o issuance of RG 1.9, Rev. 3, and tre generic letter constitute resolution of GSI B-56. The FRN will include the backfit analysis for the proposed course of action. o Staff met with CRGR on 7-25 90, CRGR Mtg 190. Slide 23 t

-O

SUMMARY

OF CRGR MTG 190 (Minners to Fralely 7/27/90) 1) RG 1.9, Rev. 3, Section C.6 will be reduced in scope. Section C.6 and C.6.1 will remain in the main body of the RG. The illustrative examples and considerations contained in Sections C.6.2, C.6.3, C.6.4, C.6.5, C.6.6 and C.6.7 will be placed into an appendix. 2) The 50.54(f) letter will be revised to more clearly note that NRC is requesting a response as to whether licensees plan to implement guidance 4 O provided, or to identify action = utiiitie= pian to Implement for monitoring and maintaining EDG reliability and to identify the type or reliability program to be used in support of maintaining required levels of performance. 3) A formal reply will be prepared in response to NUMARC'S comments on RG 1.9, Rev. 3. L L Siida 24 O

o ,s 'O i RESOLUTION ACTIONS PENDING f

1. Issue RG 1.9, Rev. 3

2. Issue 50.54(f) Letter

3. Close out GSI B 56 based on items 1 & 2

4. Issue FRN which contains Backfit Analysis s

l-i L 4 O-Siioe 25 -}}

ML20056B022

Text

SUMMARY

Navigation menu

Search