![]() |
Search Options | |||
Index | Site Map | FAQ | Facility Info | Reading Rm | New | Help | Glossary | Contact Us | ![]() |
Official Transcript of Proceedings
NUCLEAR REGULATORY COMMISSION
Title: Advisory Committee on Reactor Safeguards
Materials & Metallurgy and Plant Operations
Joint Subcommittee Meeting
Docket Number: (not applicable)
Location: Rockville, Maryland
Date: Wednesday, June 5, 2002
Work Order No.: NRC-418 Pages 1-496
NEAL R. GROSS AND CO., INC.
Court Reporters and Transcribers
1323 Rhode Island Avenue, N.W.
Washington, D.C. 20005
(202) 234-4433
UNITED STATES OF AMERICA
NUCLEAR REGULATORY COMMISSION
+ + + + +
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS
(ACRS)
+ + + + +
MATERIALS AND METALLURGY
AND
PLANT OPERATIONS SUBCOMMITTEES
+ + + + +
VESSEL HEAD PENETRATION CRACKING
AND RPV HEAD DEGRADATION
+ + + + +
WEDNESDAY,
JUNE 5, 2002
+ + + + +
ROCKVILLE, MARYLAND
+ + + + +
The Subcommittees met at the Nuclear Regulatory Commission, Room T2B3, Two White Flint North, 11545 Rockville Pike, at 8:30 a.m., F. Peter Ford, Co-chairman, presiding.
SUBCOMMITTEE MEMBERS PRESENT:
F. PETER FORD, Co-chairman
JOHN D. SIEBER, Co-chairman
GEORGE E. APOSTOLAKIS
MARIO V. BONACA
THOMAS S. KRESS
GRAHAM M. LEITCH
VICTOR H. RANSOM
STEPHEN L. ROSEN
WILLIAM J. SHACK
GRAHAM B. WALLIS
ACRS STAFF PRESENT:
MAGGALEAN W. WESTON, Staff Engineer
ALSO PRESENT:
STEVE BLOOM, NRR
KEN CHANG, NRR
STEPHANIE COLLIN, NRR
JAY COLLINS, NRR
ALLEN HISER,NRR
ANDREA LEE, NRR
STEVE LONG, NRR
MICHAEL MARSHAL, NRR
SIMON SHENG, NRR
DWIGHT SNOWBERGER, NRR
ALSO PRESENT (Continued):
KEITH WICHMAN, NRR
NILESH CHOKSHI, RES
BILL CULLEN, RES
EDWIN HACKETT, RES
DEBBIE JACKSON, RES
MARK KIRK, RES
SHAH MALIK, RES
BILL BATEMAN, NRL
JACK GROBE, NRC/R III
GIOVANNA LENGO, OGC
JENNIFER UHLE, DCM/RAM
KURT COZENS, NEI
MICHAEL LEISURE, FENOC
DAVID LOCKWOOD, FENOC
STEVEN LOEHLEIN, FENOC
PATRICK McCLOSKEY, FENOC
MARK McLAUGHLIN, FENOC
JIM POWERS, FENOC
ROBERT SCHRAUDER, FENOC
KEVIN SPENCER, FENOC
STEPHEN FYFITCH, FRA-ANP
JOHN HICKLING, EPRI
CHRISTINE KING, EPRI
ALSO PRESENT (Continued):
STEVE HUNG, Dominion Engineering
GLENN WHITE, Dominion Engineering
NATHANIEL COFIE, Structural Integrity Assn.
PETER C. RICCARDELLA, Structural Integrity Assn.
MICHAEL LASHLEY, South Texas Project
LARRY MATHEWS, SoNuclear
DICK LABOTT, PSEG
CHARLES BRINKMAN, Westinghouse
THOMAS B. HENRY, Toledo Blade
DANIEL KOFF, Cleveland Plain Dealer
JACK ROE, Scientech
ALTHEIA WYCHE, SERCH Licensing/Bechtel
C-O-N-T-E-N-T-S
PAGE
Introductory Remarks, F. Peter Ford 6
Status of NRC Bulletin 2001-01 Review, Allen
Hiser 9
Status of NRC Bulletin 2002-01 Reviews, Andrea
Lee 17
Status of MRP Work on Technical Issues, Larry
Mathews 60
Presentation of John Hickling 90
NRC Assessment of Davis-Besse Martin, Glenn
White 195
MRP Inspection Plan, Peter Riccardella 248
NRC Assessment of Davis-Besse Margin, Mark
Kirk 309
Presentation of Nathaniel Cofie 339
Status of Activities at Davis-Besse, Jim
Powers and Nathaniel Cofie356
NRC 0350 Panel Activities, Jack Grobe392
Lessons Learned Task Force, Edwin Hackett 418
Management by Leakage Detection, Allen Hiser 435
P-R-O-C-E-E-D-I-N-G-S
(8:30 a.m.)
CO-CHAIRMAN FORD: I'd like to get started please.
The meeting will now come to order. This is the meeting of the ACRS Joint Subcommittees
on Materials and Metallurgy and on Plant Operations.
I'm Peter Ford, Chairman of the Materials and Metallurgy Subcommittee. My Co-chair
is Jack Sieber, Chairman of the Plant Operations Subcommittee.
The ACRS members in attendance are everybody apart from Dana Powers. They are
George Apostolakis, Mario Bonaca, Thomas Kress, Graham Leitch, Victor Ransom,
Stephen Rosen, William Shack, and Graham Wallis.
The purpose of this meeting is to discuss the vessel head penetration cracking
and RPV head degradation issues. We've had a number of full committee and subcommittee
meetings on these issues.
Ms. Maggalan Weston is a cognizant ACRS staff engineer for this meeting.
The rules for participation in today's meeting have been announced as part of
the notice of this meeting, published in the Federal Register on May 21, 2002.
A transcript of the meeting is being kept and will be made available as stated
in the Federal Register notice.
It is requested that speakers use one of the microphones available, identify
themselves, and speak with sufficient clarity and volume so that they can be
readily heard.
We've had no written comments from the members of the public regarding today's
meeting.
The last letter that we wrote on this subject was in July 2001 in which we supported
the issuance of the Bulletin 2001-01. In that letter and in the subsequent meetings,
we raised a number of technical questions.
In his reply to the July letter, the EDO stated the answers would be given to
us in early 2002. We requested that data be presented today to support the conclusion
relating to three basic questions:
One, what do we know about the degree of degradation of the vessel head assemblies
and what is the future predictions?
Second, what are the safety issues?
And, thirdly, what are the mitigation plans?
We shall not be discussing safety culture and impacts on the reactor oversight
process as associated with, for instance, Davis-Besse, at this particular meeting.
Jack, do you have any comments?
CO-CHAIRMAN SIEBER: No, I don't.
CO-CHAIRMAN FORD: Before we proceed, Mag, do you have a statement?
MS. WESTON: Yes, one little housekeeping issue. We're going to be using the
full committee books today, Tab 2. This is the same material that's for your
book tomorrow. That's why you have your books, and I think I have opened them
all to Tab 2.
That is all of the information that I have that you have not received in hard
copy.
CO-CHAIRMAN FORD: We will now proceed with the meeting, and we will begin with
Bill Bateman of the NRR, who will make some opening comments.
MEMBER WALLIS: There is no Tab 2.
MS. WESTON: Yes, Tab 2 is turned.
MEMBER WALLIS: It's not labeled as Tab 2. Oh, excuse me.
CO-CHAIRMAN FORD: Bill.
MR. BATEMAN: Okay. While your looking for your Tab 2s, which --
(Laughter.)
MR. BATEMAN: It's a pleasure to be here today. We have an ambitious schedule,
as you can see. We're scheduled to go until six o'clock, and everybody on my
staff hopes that we're finished by six o'clock. So that is certainly our goal
and we'll do our best to get us all through by then.
And so we are looking for an interactive session. We think we're on the right
track. We hope to get some good feedback from you folks today.
I know one of the things that Dr. Ford has commented on a number of times is
the lack of data. I think this time we'll have more than enough data for you
folks to chew on.
So why don't we get started? And I guess that would be Allen Hiser.
MR. HISER: Good morning. I'm Allen Hiser with Materials and Chemical Engineering
Branch at NRR.
What I want to do this morning, very briefly to keep us ahead of schedule, is
to provide a status of the review of responses to NRC Bulletin 2001-01, which
was entitled "Circumferential Cracking of Vessel Head Penetration Nozzles."
We were here two months ago and provided a more detailed status with putting
the inspection results in the overall context so that hopefully you had some
understanding of how the data or how the inspection results are falling in line.
At this point, I want to do just one slide to give you a brief overall status.
There are no new inspection findings since the April 2002 meeting and presentation.
MEMBER WALLIS: You mean nothing has been done or nothing has been found?
MR. HISER: Nothing has been found. There have been some inspections that have
not identified any cracking or leakage.
MEMBER WALLIS: Then there are findings if you found nothing.
MR. HISER: Correct.
The MRP did make a presentation to the staff in late May with a proposed inspection
plan. I know that is on the agenda for later this afternoon, hopefully after
noon.
The NRC staff is considering a generic communication that would address interim
guidance for nozzle and vessel head inspections.
We will talk a little bit this afternoon on some of the concepts and ideas that
we have on that. No details at this point, but just some of the concepts that
we have at the present time.
In addition, we do have interactions ongoing with the industry that will provide
the technical basis for the NRC staff to develop long-term inspection requirements.
There are also activities that are ongoing within the appropriate ASME code
groups.
So that is basically what I wanted to say about the status.
CO-CHAIRMAN FORD: Is there going to be any more discussion on any of those bulletinized
things, Allen?
MR. HISER: I believe two, three, and four will have -- we will have some ideas
on later, some presentations this afternoon.
CO-CHAIRMAN FORD: So we will have some heads-up on what the generic communication
will entail? For instance inspections?
MR. HISER: A lot of concepts, more at the concept sort of a level.
CO-CHAIRMAN FORD: Will you be discussing the degree of completeness of visual
inspections versus 100 percent volumetric inspections?
MR. HISER: We can talk about that this afternoon.
CO-CHAIRMAN FORD: Okay. If you're not going to cover any more on the first bullet,
I do have a question on it which was brought up by Dana Powers at the last meeting.
You've got this famous curve -- I've almost forgotten -- of the time since the
CONY (phonetic) versus the vertical axis showing -- and I must admit to myself
some degree of conviction that the simple algorithm that we have, prioritization
fusion algorithm, seems to be reasonable.
However, Dana Powers brought up at the last meeting in April the statistical
relevance of that, given the same number of inspections have not been made at
a given time period.
Can you -- and I don't know if you remember that question. It was towards the
end of the meeting. Do you have any comments?
MR. HISER: Well, clearly the level of inspections that have been performed throughout
that, the plants listed on that chart, are different. The plants that have --
that would seem to provide the greatest support, you know, the plants that have
identified cracking and leakage have tended to have the more intensive inspections.
There are very few plants outside of that area that have done under the head,
volumetric type of inspections that are capable of detecting cracking. So many
of those plants have no results because visual exams have not identified any
leakage.
That doesn't mean that there is no degradation ongoing. It just means that thus
far, the degradation has not progressed to the point that there are leaks apparent
on the head.
We will -- what I will do this afternoon is provide a little more information
on which plants have performed which kind of inspection.
CO-CHAIRMAN FORD: Okay. Good.
MR. HISER: Maybe that will put those results in greater context.
CO-CHAIRMAN FORD: Will you also be discussing later on the completeness of that
prioritization algorithm or has it served its purpose as of now?
I am referring specifically to the facts that other countries, France specifically,
have got much more elongated prediction algorithm, taking into account micro-structure,
stress, and position of the nozzle, et cetera.
Are the NRC or the industry planning on developing such a more complete prediction
algorithm?
MR. HISER: I think that would be the -- from a technical standpoint, that would
be the desire of both the NRC and the industry. I think one of the issues that
the industry has run into in trying to put together a more thorough model is
the lack of information in some areas.
I know that in the early mid-'90s, some of the initial modeling tried to incorporate
some material parameters. And I think the results over the last several years
have demonstrated that those modeling efforts were really not as successful
as the current model appears to be.
So I'm not sure. Maybe the industry folks can speak to their efforts later during
their presentation.
CO-CHAIRMAN FORD: Can we jump -- are you going to, Larry?
MR. MATHEWS: I don't think we were planning on addressing it specifically. Basically
the model we've got right now is very simple, like you say. It has tended to
sort of mash the data that we're seeing coming in from the field.
To gather -- one of the problems is the welds. Some of the flaws have been in
the welds and it is very difficult to quantify the material and all that, properties
from a weld material.
I guess our plan was we are going to track data by fabrication and fabricator
and things like that as we do inspections. And if we start to see a demarcation,
then we can try to take that into account.
But so far, we haven't got enough data on individual heats and individual penetrations
to start to try to make that demarcation. So at this point in time, we don't
have concrete plans to do more than track the data as we get inspections over
time.
CO-CHAIRMAN FORD: Okay.
MR. HISER: And it may be, as well, that one of the biggest parameters would
be residual stresses. I think there is the variability from plant to plant and
uncertainties in that may tend to --
CO-CHAIRMAN FORD: But there's a generic relationship between fit-up angle and
residual stress and, therefore, position that you might expect as a secondary
variable.
MEMBER BONACA: I would like to ask a question. I don't need an answer now, but
I would like to understand by the end of the day why visual inspections is acceptable
as a means of detecting this degradation process for RCS. Why we would not accept
leakage in other location of the RCS as a means of detecting cracking?
And so I would not understand why in this case, it is acceptable. Is it because
of the difficulties in these inspections? Is it logical, however?
The second point -- so this is an issue I would like to understand -- the second
point is I'm concerned about the projection curve, the predicting curve that
you are showing. You are throwing on the curve MISDON 2 (phonetic), for example,
that perform volumetric inspections. They found cracks, but they didn't have
any leakage.
Therefore, you are mixing together visual results with indications from volumetric
and that creates confusion, in my judgement, about that predicting curve, and
I would like to understand why you are doing that.
MR. HISER: Yeah, I will try to clarify that this afternoon.
Your first question about visual inspections is also addressed in, I believe,
the second to last presentation on the use of leak detection as an appropriate
management tool.
MEMBER BONACA: If it is appropriate, why wouldn't it be appropriate for cracks
in nozzles or, I mean, why don't we wait until we see leakage before doing anything
to these plants? Why would we want to attend field inspections?
Thank you.
MR. HISER: If we need to make a distinction, we will do that in that presentation.
I'm not sure that will be necessary.
And with that, I will turn it over to Andrea Lee on Bulletin 2002-01.
MS. LEE: I'm Andrea Lee from the Materials and Chemical Engineering Branch,
and I'm the lead for Bulletin 2002-01 on RPV head degradation and the rest of
the reactor coolant pressure boundary.
With regard to background on Bulletin 2002-01, it was issued March 18 to all
PWR plants, and within 15 days, we asked licensees what kind of inspections
have you done in the past to identify RPV head degradation. With those inspections,
what's the ability of those inspections to determine head degradation?
In addition, after we got information on the actual inspections, we asked: what
kind of deposits, descriptors such as was it residue or staining or what types
of deposits; did you see what was left on the actual reactor pressure vessel
head?
After we asked what was done in the past and what kinds of inspection results
were obtained, we asked what kinds of plans do you have for the future to enhance
inspections or what kinds of inspections, do you have planned to address this
problem.
MEMBER WALLIS: So you asked what they were doing. Was there any kind of instruction
as to what they should be doing?
MS. LEE: It was an information request: what kinds of things have you done;
what have you seen; what are your future plans?
I will get to, in a little bit later, calls that we have had, conference calls
to address the types of things that they've seen; taking experience we've had
from talking with each of the licensees; making suggestions on how some of the
licensees we've talked to could improve based on what we have heard from other
licensees.
MEMBER WALLIS: But it is very much up to them. If you think of Davis-Besse,
until they almost accidently found they had a problem, they would have reported
everything was fine.
MS. LEE: Un-huh. I think there's been --
MEMBER WALLIS: It just up to them.
MS. LEE: I think there's been a lot of lessons learned from both on the industry
side and on the NRC side with this interaction with the rest of the 68 plants,
as well as Davis-Besse. Those types of exchanges have occurred during conference
calls.
And each of these conference calls and subsequent supplements have been put
on the NRC external Web site so that the public is aware of the kinds of conversations
that we've had.
MEMBER WALLIS: Okay.
MS. LEE: After we asked what you've done; what you plan to do in the future,
we also ask for the basis of continued operation. How can plants ensure that
they met the regulatory requirements with regard to this issue?
There was also 30-day and 60-day responses to this bulletin. The 30-day responses
are what are your inspection results in a detailed fashion; what kinds of things
have you seen, and we have asked for documentation so that the record is clear.
And then the last of the responses was a 60-day response asking what have you
done for the rest of the reactor coolant pressure boundary.
MEMBER APOSTOLAKIS: Do the resident inspectors know all of this?
MS. LEE: No, no, the inspection results--
MEMBER APOSTOLAKIS: The inspectors, the NRC resident inspectors; can they answer
these questions?
MS. LEE: In a lot of cases -- there was a TI written for Bulletin 2002-01 and
these are primarily the same inspections. So what we have tried to do in our
interactions with the plants is to make sure resident inspectors and regional
inspectors are on the actual calls.
In some cases we have gotten information that has helped to guide our interactions
with the licensees. We have gotten some good insight from the resident inspectors:
where to focus our area and focus the question.
So they are involved and they have been able to provide information.
MEMBER APOSTOLAKIS: I mean, if you ask them instead of the licensee to answer
these questions. would they give you good answers or they really don't know?
MS. LEE: Well, one example when we had a pre-qual. with the region to give us
information, the same type of questions, there were additional issues that were
raised that we were able to talk to the licensee; whereas the licensee addressed
it, but we had more of an informed conversation because we were able to dig
a little deeper from the inspector's results.
So they have been able to give us information that has helped guide the calls
and the interactions.
MR. BATEMAN: This is Bill Bateman from the staff.
In answer to that question, I don't think we were prepared to speak for every
resident inspector as to whether or not they could specifically answer these
questions if asked.
MEMBER APOSTOLAKIS: Is it up to them? I don't understand that. Your answer implies
that it is up to them to decide whether to know or not. Aren't there any rules
as to what they're supposed to know?
MEMBER BONACA: Well, the resident inspector can go every morning to the morning
meeting and listen to what the results of all the inspections. I mean he has,
right, hands-on on everything that takes place in the plant.
MR. GROBE: This is Jack Grobe from Region 3.
I think the residents would have a cognizance of licensee activities in this
area. They wouldn't have direct knowledge necessarily of the results of the
head inspection because they wouldn't have been involved in those through the
past refuel outages necessarily.
So they would assist NRR in focusing activities based on their cognizance from
being aware of licensee activities.
MEMBER WALLIS: Are they so far removed? I mean, can't they actually demand to
see photographs of what was seen instead of relying on what somebody said they
saw?
MS. LEE: Well, they have. We have seen videotapes that were provided. They've
seen pictures. There is interaction in that respect, both still pictures and
videotapes. And those are primarily how we've gotten some additional information
that has helped us focus our calls.
MR. HISER: Yeah, I think part of the confusion may be that plants that have
done inspections since last summer,- the residents, the regional inspectors
are very familiar with the results, the findings, the condition of the head,
things like that, and what sort of inspection was done.
Plants that have not had a refueling since the issuance of the Bulletin 2001-01,
the head was not a major focus area. And I think there is much less detailed
information available or detailed knowledge by the inspectors for those plants.
That's maybe where the dichotomy is occurring right now at the present time.
MEMBER BONACA: But for the boric acid corrosion prevention program, you don't
have to make an inspection of the head alone. You have other symptoms you are
looking for.
And one question would have been: in that 60-days, have they performed a lock-down
and containment or checked some for deposition, boron deposition upon surfaces?
Have they checked filters?
There are elements that can be checked even without a direct inspection of the
head.
MS. LEE: One of the things we have covered in the calls for the 15-day responses
is information known as 2002-13 which talks about containment error, radiation
element fowling (phonetic).
Since that information has come out, licensees have addressed directly on phone
calls, in some cases prompted by questions and in some cases on their own, what
they're doing to look at filters and things like that in terms of fouling.
MEMBER BONACA: But it seems to be like more -- you know, I mean, some of them
are addressed verbally, some in writing. Why can't we be more specific and request
specific answers to questions on this?
MS. LEE: Well, we have done that.
MEMBER BONACA: So that you have consistent answers.
MS. LEE: Every time we have calls, we ask for supplements to the actual response.
MEMBER BONACA: Okay.
MS. LEE: Both our written telephone conference summary and their supplement
goes on the NRC external Web.
With regard to the 15-day responses, we received all responses except for Davis-Besse.
In getting to the punch line first, we haven't identified any plants that have
the same conditions of degradation as Davis-Besse.
And the way we came to the conclusion was a priority categorization scheme for
contacting plants. This scheme was basically a subjective categorization by
the plant to guide us in how we were going to contact licensees, and it was
based on needing more information from the actual submittals.
In some cases we didn't feel we had enough clarification or verification with
what was provided. And those plants reached a higher level of priority in terms
of how -- the order that we were going to contact them.
CO-CHAIRMAN FORD: Surely, the prioritization would be exactly the same as the
cracking prioritization out of them because cracking is a precursor to the low
alloy steel corrosions; are they not?
MS. LEE: No. It's not exactly the same as the industry prioritizations for a
couple of different reasons. One is if we read a response and it wasn't clear.
For example, significant deposits were left on the head, and by significant
something that would preclude seeing the bare metal, or if there were leaks
external to the insulation such as Conoseals or canopy seal leaks, and it wasn't
clear that those were repaired within the same outage that they were found.
Those types of considerations that went into this priority scheme. So it's not
exactly the same as the industry cracking scheme.
MR. HISER: The other thing that maybe is counter-intuitive is that the plants
that have the highest susceptibility to cracking have already done head exams.
I think generally, every plant has looked at the head. So they have been able
to identify the absence of that sort of degradation.
CO-CHAIRMAN FORD: Now, when you say "look at the head," Allen, do
you mean using what technique? Purely visual or --
MR. HISER: Well, uh --
CO-CHAIRMAN FORD: That tells you nothing.
MR. HISER: They have looked visually. So they have been able to generally to
see the interface around every nozzle. If there is significant degradation or
probably -- I don't want to put a threshold. If there's degradation of a certain
level, they would have been able to identify it previously.
So the plants that are the most susceptible to cracking have been, I think,
in the best position to address this issue.
What is not readily apparent is the plants that have a lower susceptibility
maybe have not done as extensive a visual examination of the head or maybe have
not had an outage since over the last year. Those plants, we've had to rely
more on photographs and other prior inspection results.
CO-CHAIRMAN SIEBER: I guess the example there, your top priority plant is Beaver
Valley 1. In the chart of results from the bulletin that NRR compiled, the result
was called "other." When I looked into that, "other" meant
did the visual inspection, found what they interpreted to be old Conoseal leakage;
the Conoseals had been repaired; that they didn't clean the head, and so there
was residual boric acid crystals on the head, which they claimed came from the
Conoseals. They didn't clean it because of ALARA considerations.
So that would -- and their response really wasn't all that clear as to, number
one, whether they could have seen leakage from the nozzle; whether the leakage
that they saw was really Conoseal leakage; and, third, did they return the head
to a condition where visual inspection could be done unimpeded by deposits.
So maybe that helps.
MS. LEE: And in that specific case with Beaver Valley, there were subsequent
interactions with the licensee supplements to the response and future commitment
to address those issues.
So that's how that particular issue was resolved.
MR. HISER: And I think Beaver Valley was a case, as well, where we got a lot
of input from the resident inspectors and the regional staff who were on site
when they were doing the visual exam and could provide a little bit of context.
You know, if somebody says, "There were deposits on the head and we left
them there," six months ago that would have been a benign observation.
Now, the context is a lot different.
And maybe it really is a thin layer, you know, a crystal thick or something
like that. That's the kind of context that we have had to follow up on extensively
with a lot of these plants.
CO-CHAIRMAN SIEBER: I would point out that Beaver Valley 1 is on the susceptibility
list as a medium point, whereas, on the questioning list was a number one priority.
MS. LEE: Un-huh.
MR. HISER: And partly because they acknowledged leaving deposits on the head
and they were moderate susceptibility to cracking. That doesn't mean the cracking
is unlikely. I think, to the contrary, it's likely that they may have cracking.
It may be unlikely that it's through wall at this point, but there's a certain
probability that it could be.
That combination is what we saw at Davis-Besse. And maybe not -- there may not
be a scaling necessarily, but the bulletin was really focused on the conditions.Boric
acid on the head and some probability of nozzle cracking were the two main parameters.
MS. LEE: And one of the follow-on items with a number of these plants is commitment
for future cleaning. Whereas the sensitivity may not have been there at the
time, the inspection was done before leaving even, was considered insignificant
deposits. The sensitivity is there now. The next time they go in, even those
deposits will be cleaned off of the head.
So that's one of the things that has come out of these interactions.
CO-CHAIRMAN FORD: I guess I'm still missing a key point to this rationale here.
The main point of this bulletin is, to put it in layman's terms, is to make
sure we don't have another Davis-Besse sitting out there.
And your inspection method or an allowable inspection method is just to see
whether there's boric acid crystals on the top of the head. That's an allowable
measurement.
MS. LEE: No, that's not the --
CO-CHAIRMAN FORD: I can't see how that tells you anything at all about the degree
of degradation of the low alloy steel head.
MS. LEE: That's not the only parameter that the bulletin deals with. The first
few questions asks about inspection methods and how to ensure that you don't
have this particular issue. So those inspection methods go toward the 101 issue
of cracking.
CO-CHAIRMAN FORD: When they answer the question to give you a rationale on why
they should continue to operate, do you accept the rationale that I haven't
seen any boric acid on my head and, therefore, I have no problem?
MS. LEE: No. It's a combination both of what have you done inspection-wise to
see -- because axial cracks come into play with this phenomenon. It's not just
the 0101 concern of circumferential cracks above the nozzles.
So the first part, it's almost a combination of the two bulletins. The first
part asks have you done inspections; what types of inspections have you done
of the nozzles to see what you actually have in regard to cracking and degradation.
The second part asks about what have you seen in terms of deposits on the head
and things like that. So it is a combination of the two issues and the two concerns.
MR. HISER: But I think, fundamentally though, if I have a head and I'm worried
about degradation and I go and look at the head and see no degradation; I can
check that plant off.
CO-CHAIRMAN FORD: It didn't at Bouget 3.
MR. HISER: Now it may be that there are conditions that could lead to degradation,
but at the present time I do not have degradation ongoing.
CO-CHAIRMAN FORD: But my point is at Bouget 3, for instance, there was no boric
acid and yet there was cracking.
MR. HISER: Right.
CO-CHAIRMAN FORD: So that's one. What we need is one more, guys, and we're dead.
MEMBER BONACA: I think, in the context of this question, I agree with that.
Even for the plants that already perform inspections, clearly when they did
inspections, they did not know that Davis-Besse would occur.
There are tell-tale signs that Davis-Besse was seen. For example, you know,
the bottom-up spraying that they have identified and never figured out why they
had it, but I'm sure that Oconee didn't look for it whenever they did an inspection.
I think it would still be wise for them to go back to the inspections, review
what they did, try to remember if there were signs that Davis-Besse had identified
as delta signs.
So I think just the fact of having inspected visually those heads, in the very
difficult conditions of the inspections, many of them, I don't think should
be just sufficient. I think that they should look back at what they did and
try to interpret some other signs that may have seen.
MR. HISER: Well, I think given the information that is in the 2002-13, many
of the responses address those kind of indirect indicators directly. They said
-- and I think you were at a plant -- they had a used filter from their radiation
monitors and a brand-new filter. They were indistinguishable.
Plants are looking at those kinds of indirect indicators as well. If I look
at the head and see no degradation, that gives me a good feeling right now that
I do not have a Davis-Besse situation at that plant.
Now, that doesn't tell me in five years I may not, and then it becomes incumbent
on the NRC and the industry and the licensees to implement effective inspection
programs to ensure not that we don't get Davis-Besse, but that we don't get
down the road any of the precursors that led to Davis-Besse. That's the thing.
MEMBER WALLIS: You are very reassuring. I mean the crack growth varies by orders
of magnitude and the graphs that we look at -- you are going to say that just
because someone didn't see some crystals, that there is no crack there which
isn't going to grow more rapidly than you thought and is going to lead to some
incident?
MR. HISER: I'm saying right now we don't believe that the conditions are there
at any plant.
MEMBER WALLIS: I have the concern with this slide. I mean, the statement that
was made was a little more reassuring than the first one that the staff has
not identified. I think it was a little different. It was ten minutes ago. I'm
not particularly remembering it, but I think you wanted to reassure us that
there wasn't another Davis-Besse out there.
The fact that the NRC has priority categorization doesn't have any effect on
the physics. The fact that the NRC has no concern about 49 plants really surprises
me.
There has to be concern about every plant out there.
MS. LEE: Yeah. With regard to the no concern bullet, that doesn't mean we don't
have clarifying questions or verification questions. So some of those plants
do have questions associated with them. It was prioritized as no concern just
based on the order of --
MEMBER WALLIS: Or what you thought you knew before Davis-Besse.
MS. LEE: No, actually this was after, after Davis-Besse.
MEMBER WALLIS: After Davis-Besse?
MS. LEE: Yeah.
MEMBER WALLIS: After Davis-Besse, you had no concern for 49 plants?
MS. LEE: The no-concern categorization really was based on -- and the whole
priority scheme is based on -- the order of contacting. And it is caveated to
say we may still have clarifying questions or something that wasn't particularly
clear.
MEMBER WALLIS: So when something was found at one of these 49 plants, someone
is going to remember that the NRC had no concern.
CO-CHAIRMAN SIEBER: I gather the no concern means you didn't have a concern
about the response to the original bulletin.
MS. LEE: Yes. There may have --
CO-CHAIRMAN SIEBER: It doesn't mean that you didn't find something.
MS. LEE: It doesn't mean no concern with the actual what would be future occurrences
at the plant. It was based on the licensee response was primarily complete,
but there may still be a question here or there; to just ensure that we have
all the information we need to make the informed decision.
CO-CHAIRMAN SIEBER: I guess one thing that bothers me is the fact that you can
have nozzle cracks and you can't find them by visual inspection unless they
are through through wall and leaking. To me, that gives me little comfort.
MEMBER BONACA: That is exactly right.
MR. HISER: We'll talk about that later this afternoon, but I think --
CO-CHAIRMAN FORD: It's important line right now.
MR. HISER: The purpose of this bulletin was really short term. Do we have similar
conditions at any of the other 68 plants?
The bulletin has served its purpose in that we don't think there are any other
plants out there with those conditions. Now that doesn't tell us that in two
years something could not develop because clearly it could.
But, for the present time, we don't think that's the case. We are working to
implement inspections that will ensure that in two years we can come back and
say this problem is being managed and will be for the long term.
That's where we are after. The bulletin is just a short-term instrument to give
us a status report on where plants are with this degradation.
CO-CHAIRMAN FORD: You're saying here that you don't have any situations similar
to Davis-Besse, based apart from the first ones issued, 100 percent volumetric;
based primarily on visual. And that makes me feel really worried.
Because we may hear later today about what the specific criteria or design criteria
that will give you the local annulus environment, that could give you one inch
per year low alloy steel corrosion rate. We may hear that later on today.
But until we hear something definite, some design feature that would preclude
that you've got to assume --
MEMBER BONACA: Yeah, until you rely on the visual, I mean, you know, certainly
you know that as soon as a crack develops, it could start the process of erosion
and corrosion of the head.
MR. HISER: I think that --
MEMBER BONACA: So we need to hear more about it.
CO-CHAIRMAN FORD: Are we missing something, Andrea? Visual -- you, as professionals,
are sure that by looking on the head and not seeing boric acid, therefore, you
do not have low alloy steel corrosion.
MR. BATEMAN: This is Bill Bateman from the staff.
I would just like to refresh everybody's memory of the process here. We issued
a bulletin requesting information from the licensees with the respect of the
condition of their head. We got 68 responses 15 days after we sent the bulletin
out.
Those responses were under oath and affirmation. The licensee knows the condition
of their heads much better than we do. And so we basically believe what they
tell us in their responses.
What we have been doing since then, which is a little over two months ago, is
having phone calls with licensees in a priority order here based on the quality
of their response and trying to get a comfortable feeling; fill out the details
that are missing; et cetera, to come to some kind of conclusion with respect
to whether or not we feel they have the potential for the problem.
We have not looked, personally, at any of these heads. Well, maybe I'll take
that back. Maybe we've looked at one or two heads. But again, the licensees
have sent us their response under oath or affirmation and they basically have
made the claim, each and every one of them, that they don't have any evidence
of something similar to Davis-Besse.
So that's the process we're in.
I think to take the staff to task for not having seen each and every head and
making a visual observation is not fair.
CO-CHAIRMAN FORD: Obviously, you can not go in -- you personally can not go
and look at every head. I just -- I'm trying to delve into the rationale.
Right now what you're saying is essentially engineering judgement. You feel
comfortable by engineering judgment based on --
MR. BATEMAN: We feel comfortable based on the licensee responses that came in
under oath or affirmation, the descriptions that they put into those responses,
asking them questions in a priority order of those licensees who didn't give
us enough information so that we could come to a clear conclusion. Yes, we feel
comfortable based on that.
Their responses and our subsequent questioning of their responses and this kind
of a priority order.
MR. HISER: The two things that I guess I would add to that is if you do have
corrosion ongoing, you do have water leaking, you do have boric acid, that goes
somewhere. The corrosion products have a much lower density than the low alloy
steel.
It's going to be obvious somewhere that something is going on. If you look back
at the Davis-Besse visual examination results from their head, there were many,
many, many signs on the head, containment air coolers, radiation monitors, that
something was going on.
These other plants have not identified any of those kinds of indicators that
we think are persuasive in indicating that there is no degradation going on
in this area.
MS. LEE: And in many cases --
MEMBER LEITCH: What concerns me is that on November 9, we met here and were
being briefed on the results, the early inspection results from or the early
responses from Bulletin 2001-01, and one of the things that we were told at
that time was that Davis-Besse, in trying to justify why they didn't have CRDM
cracking at that time, referred to some earlier video tapes they had done of
their head.
They did videotapes in 1996, 1998, and 2000. They claimed at that time that
they were not specifically looking for CRDM cracking, but they were trying to
use those tapes as a justification for why they didn't have CRDM cracking.
But they further claimed that they made those videotapes specifically looking
for head degradation as a result of boric acid on the head probably from historic
flange leaking and claimed that after reviewing those video tapes, from those
three inspections, they were satisfied that there was no head degradation.
And so my question is: aren't we hearing the same thing from these plants?
In other words, when we probed deeper then into the Davis-Besse situation, there
were questions about how, well, we couldn't see certain CRDMs very well.
I mean, is there anything here about how-- what percentage of the head they
can really look at? And how does one interpret what is seen on the videotapes?
Is it what is referred to as "popcorn"? Is it what is referred to
as "lava"? Is there common understanding when someone says "popcorn"
and somebody else says "lava"? Do we really know what we're talking
about there? If somebody talks about "white deposits," "red deposits"
-- I mean, there is a lot of subjectivity in those kind of words.
MEMBER BONACA: Furthermore, a number of these plants have never inspected their
head, I would suspect. I mean some of the 49 plants are not concerned. They
may not look at them.
MEMBER LEITCH: So what I'm saying is in the time frame of November 2001, Davis-Besse
would have satisfied these criteria, not only could have, but did effectively
answer this bulletin before it was written in response to questions at this
meeting and answered them in a way that satisfied us all, and we were wrong.
MR. HISER: I would expect if you look at their root cause analysis report, that
I think some of the information provided in there is not necessarily consistent
with what the ACRS was told and what the staff was told last fall. That would
be the main comment that I would make in that area.
The other thing is that, again, from the input we've gotten from the residents
and the regional inspectors, from documentation, for plants that are not inspected
since prior to Bulletin 2001-01, there tends to be some photographic evidence
of the condition either of the head or the insulation that is directly attached
to the head, and if that is undisturbed, that, again, is a positive indicator
that there is nothing going on.
If you get corrosion, the products are going to go somewhere. For the short
term, that provides us with the basis for the first statement on here. For longer
term management, I don't know that that is an acceptable approach.
We'll talk about that later this afternoon. Because at the present time we are
looking for an outlier condition, you know, gross degradation. For long term
management, that's not the correct standard to use.
We want to ensure that we don't have precursors. We don't want to get -- we
don't want to say how far down the path. We don't want to be on the path, overall.
We don't want to preclude the industry from being on the path.
MEMBER LEITCH: So I guess what you're saying is the reason that I should have
some confidence in these results versus what Davis-Besse told us in November
2001 is that these results are done -- are made with an informed judgment because
we now have the history of Davis-Besse.
MS. LEE: Yeah, I think that is one of the most important distinctions to make.
When we were in the November time frame, no one could have imagined that we
would have discovered this type of degradation on a reactor vessel head.
We're in a different climate now. Because of Davis-Besse, there's heightened
sensitivity to these types of issues, and again, the plant that I visited with
regard to filter papers, and if you recall, I think it was the April 2000 picture
at Davis-Besse with the corrosion pouring out of the mouse holes onto the reactor
vessel studs. There were many, many indications of degradation on that head.
I think with the climate that we're in now, people have gone back, looked at
their pictures; have gone back, looked at inspection results; and are doing
inspections now with a more in-tuned eye and more informed decisions on what
they're actually looking for.
That's why I personally think that there's much more scrutiny in terms of per-Davis-Besse
and post Davis-Besse.
MEMBER LEITCH: But some of these plants have no new inspection results really
since Davis-Besse. In other words, they are just manipulating old data and analyzing
old data in light of the Davis-Besse incident.
In other words, a lot of this response represents not new videotapes or new
photographs, but going back and looking at videotapes and photographs previously
and interpreting them in light of Davis-Besse; is that --
MS. LEE: But they actually do have indicators. The indicators would always be
there whether they had done an inspection or not. For example, unidentified
leakage. One of the things that a lot of the plants have indicated is they have
extremely low unidentified leakage. The tech specs say one gallon per minute.
They're at somewhere like .06 gallons per minute.
So the indicators are an important factor with the rest of the plants, even
if they haven't done inspections.
MEMBER LEITCH: Well, yeah, but the difference between .1 and .2 gallons per
minute could be very significant as far as this is concerned. Your point operators
may not react to that kind of change.
What I'm saying is this is small, I think, compared with the normal variability
that one sees in unidentified leakage.
MR. HISER: Some of the things just to -- you know, how did some of the plants
come on this high priority list is an example of programmatically they did not
tell us if they had Conoseal leaks or something like that. Did they immediately
clean-up the boric acid spillage?
If they did not say that, we were asking for additional information regarding
their practices. There's a variety of practices in areas like that.
So we tried to look at the holistic approach, looking at all of the available
information from the programmatic aspects to maybe interpreting old inspection
results and any documentation of those inspections, plus the more recent inspections
that clearly have been focused on this area as a prime area of concern.
So we tried to gather all that information together to make the determination
in this case.
MEMBER LEITCH: Was one of the variables that you considered the ease with which
the head could be completely inspected?
MS. LEE: A lot of the questions we've asked licensees very directly is did you
get 100 percent inspection of 360 degrees around the circumference of each nozzle,
and in some cases the answers were we got 96 with a robotic-type crawler, but
we got the rest with a camera on a stick.
So we've gotten very specific in terms of what they could see, what they couldn't
see, and what inspection methods they actually used.
CO-CHAIRMAN FORD: Even with conformal insulation?
MS. LEE: Pardon me?
CO-CHAIRMAN FORD: Even with insulation which is conformal to the pressure head?
MS. LEE: In the cases where there is insulation, for example, glued to the head
or contoured to the head, we've had discussions about what are your plans.
In some cases, there have been nondestructive examinations performed. So --
CO-CHAIRMAN FORD: So in those cases there was nondestructive --
MS. LEE: Not in every case, but in some cases there were. In the cases where
there haven't been inspections done yet, they have plans to do that in the next
inspection.
CO-CHAIRMAN FORD: So it's not 100 percent then. In those cases where they were
not able to do a visual --
MR. HISER: But the kind of -- I think a typical situation would be, as a part
of our normal outage inspections, we look at the insulation. We have seen no
disturbances on the insulation. We've seen no staining, no deposits --
CO-CHAIRMAN FORD: Okay.
MR. HISER: -- nothing like that on the insulation. We looked at the flange every
outage.
MEMBER WALLIS: But are there indications of what those deposits would look like
on the insulation, that they would be visible?
MR. HISER: These insulation packages are pretty much watertight.
MEMBER WALLIS: This stuff is creeping under the insulation and eating away the
head and you wouldn't see anything.
PARTICIPANT: Right at the top of the insulation so you can see it.
MR. HISER: Well, they also examine the flange area. If there is anything ongoing
under the insulation, we would expect that it would flow out and be visible
there.
MEMBER WALLIS: Okay.
CO-CHAIRMAN FORD: Could I try to come to a kind of an agreed upon conclusion
as to where we are?
MEMBER ROSEN: Peter, before that, could I --
MS. WESTON: And I have a question too.
MEMBER ROSEN: -- could I make a comment?
CO-CHAIRMAN FORD: Sure, you bet.
MEMBER ROSEN: Allen, you said something I thought was very important, which
was that the root cause analysis report, presumably gave the Davis-Besse's report
was what you were referring to -- gives you different information than what
was provided to the staff and to the ACRS at various times; is that correct?
MR. HISER: That's my understanding. Just reading through some of the observations
of their inspections.
MEMBER ROSEN: I'm saying is that what your saying?
MR. HISER: Yes.
MEMBER ROSEN: I assume someone is following that up.
MR. HISER: I believe that's my understanding.
MEMBER BONACA: Was it different or was it additional?
MR. HISER: Probably additional as much as anything. Character deposits, colors,
things like that that were not -- information that we were not aware of.
MS. LEE: And also degree of cleaning the head, the level of cleaning.
MEMBER ROSEN: But I'd like to have some assurance that someone is carefully
sorting that out.
MR. HISER: Regarding Davis-Besse, I think Davis-Besse has issued press releases
to that effect, that there are regulatory activities going on.
MEMBER ROSEN: No, I don't really -- I am interested in what Davis-Besse says,
but I would prefer to hear it from the staff, that someone in the staff is carefully
sorting out what Davis-Besse told NRC and the ACRS and what they now know and
wrote down in their root cause analysis report.
MR. GROBE: This is Jack Grobe from Region III.
There are two activities that are ongoing in that regard. One is follow-up inspection
to the AIT inspection evaluating the results of that inspection which included
not what the licensee told the ACRS, but certainly what the licensee told the
staff. There's also an investigation ongoing by the Office of Investigations
into various aspects of what resulted in the head degradation at Davis-Besse.
I'm not sure it is appropriate to discuss the details of exactly what issues
the Office of Investigations is focusing on in a public forum.
MEMBER ROSEN: Thank you.
MS. WESTON: I have a question. How much information do you get documentation
to independently verify the statements that are made by the licensees? For instance,
photographs, videotapes, things like that. Does the staff actually get that
information and look at it independently to see?
And I'm thinking basically of the Davis-Besse photo that apparently had been
taken some time before it was provided. What do you do to independently verify
any of this information?
MS. LEE: In a lot of cases, the licensees have included pictures right with
their initial response and also indicate that they have videos. We have followed
up with some of the plants and asked for actual videos.
Also the residents are an important factor in that as well. Because a lot of
times, they are the first in line that have seen these pictures, seen the videos,
were with the licensees when the actual inspections were occurring. So there
is the opportunity for independent verification.
And I think in terms of Davis-Besse as Allen said, there were some differences
in what was provided back in the November and December time frame and then what
was provided after the degradation was discovered.
So again, as Jack said, there is follow-up, investigative follow-up as to sorting
out all of that and what was provided and how it differs now.
MS. WESTON: So my question, then, is how do you assure that is not happening
again?
MS. LEE: I think in terms of the information that we have gotten and the information
that we have followed up on, we try to do integrated types of reviews both --
as Bill said, it's under oath and affirmation. We go with that as the first
line.
But we've constructed the questions to dig deeper into what they've provided.
In some cases, we have asked for additional pictures, asked for additional video
and additional evaluation of that with regard to what the resident saw right
directly on conference calls. We try to sort through as much information as
we can get at the time.
MR. HISER: In at least one case there were some photos that we were provided
of the condition of the insulation, as an example. It appeared, to us, to indicate
some sort of degradation of the insulation. It wasn't obvious if it was external,
if it was from the head.
In that case, the licensee went in, removed the pieces of insulation, did a
bare metal visual exam of the head itself, and confirmed that there wasn't a
degradation at that point.
It has been a myriad of approaches to try and to reach conclusion on each plant.
But at this point, there are still some outstanding plants that we need to nail
down the final details on, but in an overall sense, we have a very good feeling
that there is not significant degradation going on.
MEMBER WALLIS: How big are these responses? Are they two pages, a thousand pages,
ten thousand?
MS. LEE: No.
(Laughter.)
MS. LEE: Did you say a thousand?
(Laughter.)
MS. LEE: No, it's not a thousand. It varies. Typically they may be like, for
example, the 15-day responses, they could be 40 pages; they could be 20 pages.
MEMBER WALLIS: So you've read 68 20-page responses?
MS. LEE: Some --
MEMBER WALLIS: So it come down to what a professor grades in one day?
MS. LEE: -- on average.
MEMBER WALLIS: The kind of thing a professor grades in one day?
MS. LEE: No.
MEMBER WALLIS: You're only 20 percent complete in a month?
MS. LEE: You're talking about the 60-day responses now? You're going down to
--
MEMBER WALLIS: Oh, am I going down to the -- am I out of -- okay. I'm sorry.
MS. LEE: Yeah. We were discussing, really -- the original discussion was on
the 15-day response.
MEMBER WALLIS: So they have a fat one so that they are much bigger?
MS. LEE: Well, the 60 -- I'll just go on to the end of the slide -- the 60-day
responses were due May 18. And we've gotten the last of them in at the end of
last month, the end of May.
The staff has begun the review. It's been about 20 percent done.
MEMBER WALLIS: Those are the fat ones?
MS. LEE: The 60-day responses are the rest of the reactor coolant pressure boundary.
MEMBER WALLIS: Oh, the rest of them, okay.
MS. LEE: And again, that varies. There are some that are 40 pages. There are
some that are less.
MEMBER WALLIS: So the 15-day responses -- I'm sorry -- have all been reviewed
thoroughly?
MS. LEE: Yes.
MEMBER WALLIS: Okay.
MS. LEE: Yes. And just another note about the 60-day responses. Some of them
may refer back to past programs on boric acid corrosion programs. So in terms
of the length of them, they may be smaller because they are referring back to
information that was provided on the docket.
MEMBER WALLIS: Isn't part of the problem in reviewing is that you allow them
too much latitude in the way in which they present the evidence?
If you were very firm about that you must have evidence of 360 degree inspection
of every nozzle -- we want to see it. We want it at a certain place in the report
-- then you could run through them all and see if there was any concern.
MS. LEE: Un-huh.
MEMBER WALLIS: If every report looks different, it is much more difficult to
review it, isn't it?
CO-CHAIRMAN FORD: I'd like to bring this one towards a conclusion.
MEMBER BONACA: Could I just make one? We talked about Davis-Besse, and I think
Davis-Besse gives us the wrong comfort in my judgment. Because however responsible
Davis-Besse will be found to be, we have to recognize we were all surprised
by the finding we had at Davis-Besse. We did not -- I did not expect -- that
kind of degradation.
Therefore, I don't think we can be comfortable about all the remaining plants
out there that are sitting with insulation on their heads expecting that what
will happen will be either what we discovered last year, that axial cracks might
become circumferential, or we will discover this year that cracks may become
degradation of the head. There may be something else that is developing there.
So I think it is important that we don't get too much comfort with the fact
that maybe Davis-Besse made some wrong judgments.
MR. HISER: I think short-term comfort is all. For today, I think we have comfort.
For the future, we need --
MEMBER WALLIS: Could I have just one quick question for the fact -- you may
have said this, but of the seven, four, and eight plants that your contacting,
what is the status of that? Have those contacts been made or are they yet future?
MS. LEE: For all of the plants and even the majority of the no-concerns plants,
the contacts have been made. The calls have been documented and the supplements
are coming in. The majority have come in, have been documented and put on the
Web site.
MEMBER WALLIS: I'm not sure I understood your answer. For the high, medium and
low priority plants, they have all been contacted?
MS. LEE: Yes. And then some of the no-concerns plants that we may have clarifying
questions on, the majority of those have been contacted as well.
MEMBER WALLIS: Thank you.
CO-CHAIRMAN FORD: Let me finish off, unless there's any burning questions, with
a -- could you keep that up please, Andrea?
I'd like to suggest that a better wording which would be a compromise wording
of the first statement there is that you have not identified any plants with
the gross lava flows that you have observed at Davis-Besse.
(Laughter.)
CO-CHAIRMAN SIEBER: However, until we do the 100 percent examination on all
plants or until we understand the chemical and geometrical aspects that would
give rise to one inch per year corrosion rates, you can't assume that there
isn't an incipient Davis-Besse out there.
Is that a fair compromise statement?
MR. HISER: At this point, we are not far down the path. What we need to do now
is make sure that nobody is on the path that would lead to Davis-Besse.
CO-CHAIRMAN FORD: Okay.
MR. HISER: I think that's correct.
CO-CHAIRMAN SIEBER: Right.
CO-CHAIRMAN SIEBER: Maybe as another sort of summary of what I thought I heard
when we complained about visual might not be being adequate enough to identify
cracking, visual was originally chosen because of fracture mechanics arguments
that say even if it leaks a little bit, it is not going to separate and go sail
on up to the roof of the containment, which I thought was okay at the time.
But that is just the first step. Sooner or later -- and you indicated it yourself
-- that you've got to move to a better inspection technique than a visual or
the camera on a stick.
MR. HISER: That's correct.
CO-CHAIRMAN SIEBER: Is that the right impression?
MR. HISER: I think that is correct. I think we will talk about that a little
bit later this afternoon, but I think Davis-Besse has raised the bar a little
bit in terms of the information that we need. How far down the path of leakage
and cracking are we comfortable with?
It may be that we need to move back quite a bit, push the bar back.
CO-CHAIRMAN SIEBER: Okay. Mr. Chairman, I now feel comfortable that we can move
on.
CO-CHAIRMAN FORD: Okay. Andrea, Allen, thank you very much indeed.
Larry, are you up?
MEMBER WALLIS: That's a new reactor design you've got there?
MR. MATHEWS: Yes, it has plenty of containment.
MEMBER ROSEN: This is some report.
CO-CHAIRMAN FORD: Larry, I understand that Glenn White wants to give a presentation
before lunch. Can you arrange, whatever you are both going to do, so we can
get Glenn in before lunch?
MS. KING: Yeah, we currently had that planned for the --
CO-CHAIRMAN FORD: Very good. Excellent.
MS. KING: -- for the two and a half hours.
MR. MATHEWS: I'm Larry Mathews, by the way, from Southern Nuclear Operating
Company and the Chairman of Alloy 600 Issues Task Group of the EPRI Materials
and Liability Program.
This is Christine King, the project manager from EPRI. We'll have other speakers
and I will go over that on the agenda here.
I have a few minutes on the status. Then we're going to turn it over to somebody
who knows a lot more about this stuff than I do. We have John Hickling from
EPRI, who will make a presentation on our Alloy 600 crack growth rate work and
the expert panel and where we stand on that.
Then we have Pete Riccardella from Structural Integrity, who will discuss the
probablistic fracture mechanics model, and also how he used that or how we used
that as the basis for our initial cut at an inspection plan.
I have just a few minutes on collateral damage.
Then Glenn White from Dominion Engineering will come up and make a presentation
on the technical assessment that we have ongoing.
Then later this afternoon, we are going to talk about the inspection plan and
where we stand on that.
CO-CHAIRMAN FORD: Will you be discussing at all during the day any work on the
physics of how you can get one inch per year, low alloy steel corrosion rate?
MR. MATHEWS: That's Glenn's presentation.
CO-CHAIRMAN FORD: Right.
MEMBER WALLIS: I guess it is chemistry, too.
MR. MATHEWS: Yeah.
CO-CHAIRMAN FORD: By "physics," I meant atom by atom.
MR. MATHEWS: Well, it's physics. Chemistry is a subset of physics.
MEMBER BONACA: What's MRP? What's MRP stands for?
MR. MATHEWS: Material Reliability Program.
This is a flow chart -- and I can't see it -- this is a flow chart of basically
the strategic plan that we have laid out for addressing the head penetration
cracking issue. We have a similar one for the VC summer type issues.
We did not include in here work on Davis-Besse. This was put together before
Davis-Besse. In fact, our initial cut at the inspection plan wasn't addressing
the Davis-Besse issue. We were saying that it should be relied upon by -- well,
we should rely on the 8805 program and improvements that need to be made perhaps
to that program. However, based on comments we got, we are going back to take
a look at what we really want to say in the inspection plan.
MS. WESTON: Larry, excuse me. Members, there is a larger version of this, page
number 17, handwritten 17 in your book.
CO-CHAIRMAN FORD: Thank you, Mag.
MR. MATHEWS: How would you get that?
(Laughter.)
MS. WESTON: Magic.
MEMBER APOSTOLAKIS: Are all of these in the book?
MS. WESTON: I'm not sure. This is from a previous presentation. I will tell
you if the page is there in the book. But you have this handout which has them,
but I have some of these duplicate slides that are in the book.
MEMBER APOSTOLAKIS: Okay.
MS. WESTON: So handwritten page 17 --
MEMBER APOSTOLAKIS: You're right.
MS. WESTON: -- under Tab 2, has this in a larger version.
MEMBER WALLIS: Is there some rationale to this figure?
MR. MATHEWS: We ultimately want to arrive at a final reactor pressure vessel
head nozzle safety assessment that would be submitted to the staff. And all
of these other things are what we're working on to flow into that, including
and what we will talk about today are the ones that are highlighted in pink
or red.
The susceptibility rankings briefly. We are going to have an extensive presentation
on the crack growth rate and the probablistic fracture mechanics in the inspection
plan later this afternoon.
MEMBER WALLIS: So it is all cracking?
MR. MATHEWS: It's all cracking on this chart. The MRP is doing work relative
to the wastage issue, and Glenn will be discussing what he has been working
on at the end.
MEMBER WALLIS: Now, two questions. First of all, this is all Alloy 600 and 182
and 82?
MR. MATHEWS: Right.
MEMBER WALLIS: Anything on 690?
MR. MATHEWS: No, not in here.
MEMBER WALLIS: Is there somewhere?
MR. MATHEWS: It's going to be looked at, yes, but we don't have it now.
MEMBER WALLIS: I ask the question because in all likelihood some of the stations
will be going to 690 Alloy 52 replacements where necessary.
MR. MATHEWS: Soon. Yes.
MEMBER WALLIS: And therefore, presumably the staff are going to ask for some
quantification of the fact of improvement.
MR. MATHEWS: Yes, and there is some information out there, and it will all be
pulled together. Ultimately, the inspection plan should be addressing what's
the right thing to do for those materials also.
MEMBER WALLIS: Which comes to my second question: what is the time line?
MR. MATHEWS: We are shooting for this in the third quarter of this year.
MEMBER WALLIS: So a lot -- most of these have been finished?
MR. MATHEWS: Most of them are very far down the road.
MEMBER APOSTOLAKIS: Was this -- this was not started because of Davis-Besse,
right?
MR. MATHEWS: No, no. This was started because of Oconee.
MEMBER APOSTOLAKIS: So coming back to Dr. Wallis' question, where do we enter
this?
MR. MATHEWS: Well, it's all parallel really. The susceptibility ranking was
the first thing that we put together. It was just the time and temperature ranking
to try and figure out what plants were most susceptible and need to be concerned.
So that was put together and I guess it was actually submitted to the staff
in response to 2001-01.
MEMBER APOSTOLAKIS: So do the colors mean anything?
MR. MATHEWS: The red means it's just what we're going to be talking about today.
This is the final product color, and they're pretty.
(Laughter.)
MR. MATHEWS: The green, I think, was stuff that we were actively working on
at that point in time when put these colors. You did the colors?
MS. KING: I did the coloring. Christine King with EPRI.
The green are things that we have interacted with the staff on.
Some issues are red here today. It doesn't mean we haven't talked to the staff
about it. It just means that we're here to talk to you guys about it today.
The yellow are things that we would like to have interactions with NRC staff
on. When we get to a risk final, put together a risk assessment, and we would
also like to talk to them about the inspection technology demonstrations that
we have been ongoing at the EPRI and DE center.
CO-CHAIRMAN FORD: When you say "would like to," Christine, this is
one of the other questions I had, is not only the timing, third quarter this
year for the blue, but at what points do you have interactions with the staff
on a down-and-dirty basis, data-to-data basis?
MR. MATHEWS: We've already had interactions on several of these crack growth
rates and the probablistic fracture mechanics. We've had -- I thought it was
a pretty down-and-dirty meeting.
(Laughter.)
MR. MATHEWS: A couple of meetings on those issues with the staff and --
CO-CHAIRMAN FORD: Okay.
MS. KING: Yeah. We've spoken to the staff a few times on crack growth rate as
well as PFM. We've been interacting on the PFM model since last September with
the staff and incorporating comments and changes.
MEMBER APOSTOLAKIS: Again, is this going to be the traditional scientist's approach
and the expert's approach to this? Or is it going to be a realistic risk assessment?
(Laughter.)
MEMBER APOSTOLAKIS: For example, if I look at this and I know what Davis-Besse
did, where would I go and say, "Well, gee, this is really where they did
things that were surprising"?
Like visual inspection guidelines, are you going to assume that these will be
performed in a way that the intended result will be, in fact, achieved? Are
you going to assume that the crack growth rates are the scientific rates, when
I read here that the B&W owner's group had underestimated those rates in
their regional calculations?
I mean, are you going to have issues like that in here? Otherwise the result
would be ten to the minus X and we pick X?
(Laughter.)
MEMBER APOSTOLAKIS: Well, I mean, at some point you have to draw the line and
say they are not doing it. The program is there, but they are not implementing
it correctly.
I know one of our issues addresses safety culture issues, but why else are we
doing this?
MR. MATHEWS: The inspection plan is going to be finalized and out to the industry.
It is my understanding that already INPO in their visits to the sites are looking
into how plants have done boric acid walk-downs, et cetera.
The inspection plan would probably ultimately be audited by the industry itself
by INPO. That would probably be the way that it would go.
MEMBER APOSTOLAKIS: Shouldn't there be other boxes with question marks inside
feeding into the risk assessment for somebody else to worry about? Or is this
the only thing that goes into the risk assessment?
It says probablistic fracture mechanics and there is the arrow to the risk assessment
which is of concern to me.
MR. MATHEWS: Everything is feeding into the risk assessment. All of it, ultimately
if you look at it, gets into that box.
MEMBER APOSTOLAKIS: But this is the material expert's review, isn't it?
MR. COZENS: This is Kurt Cozens from NEI.
And I might be able to help just a shade on this because I think I understand
what you are asking, and if I might just interject for a second.
The MRP process has an executive steering committee, and when we say executives
we are talking about the chief nuclear level. These individuals that sit on
this executive board have and do review the technical work that has been put
out by the ITG, reviewed by its own infrastructure that critiques this.
They look at this, not only from a technical issue, but from what I'll call
the policy level issue of what is the right thing to do. And I think that is
the essence of what you're looking at.
Not only what do the engineering numbers say, but is that really the right thing
to do in managing their plants?
So that is a very big consideration. I believe the staff is looking at that
from the same point of view. You know, the numbers may tell us one thing, but
when you really look at the real world, what are the things that should be accomplished?
And there is a lot of oversight at a high level within the industry to ask some
of those tough questions.
Larry, I defer that back to you. But I think, I believe that's what you were
driving to, wasn't it?
MEMBER APOSTOLAKIS: Well, these are policy.
MR. MATHEWS: The risk assessment question or risk assessment that is being done
is not just a bare bones. We are putting conservatism in there at various stages,
and you'll see some of that in Pete's discussion of the PFM work.
MS. KING: And I guess I would like to point out that this whole thing is fed
with the inspection data that we are getting from the field. We continue to
evaluate that data, what we're finding in the field, and reviewing our work.
MEMBER APOSTOLAKIS: But if I want to, I mean there is such a thing as Defense
in Depth, and the structuralist interpretation is that if I'm wrong or if I
don't have good information, I want to make sure that nothing will go wrong.
So in light of Davis-Besse now, if the inspections are inadequate or if the
crack growth rates are underestimated, what is it that is protecting me? What
Defense in Depth do I have in here that says, yeah, your estimate is ten to
the minus six, but it is really .3?
So something needs to be there to protect me and I don't see that.
MEMBER WALLIS: There's a containment.
MEMBER APOSTOLAKIS: Oh, the containment. I think we have to ask those questions
because if we don't ask them now, we'll never ask them.
MEMBER BONACA: That's why we're asking questions about the inspections. Because
if you went in now --
MEMBER APOSTOLAKIS: If things are implemented the way they are supposed to be
implemented, then I will believe this analysis. But unfortunately, sometimes
they are not.
So I have to have some measure somewhere that satisfies my Defense in Depth
needs. I don't know how we're going to do that.
MEMBER ROSEN: Well, Mario, you have more than the containment. You have your
emergency core cooling systems as well.
MEMBER BONACA: Of course.
MEMBER APOSTOLAKIS: Anyway, let's go on.
MR. MATHEWS: I was just going to show what we're going to talk about.
I'm going to turn your ranking around. Okay? What we have done and what we have
decided is the right way to look at this thing in the future is not to try and
take a reference plant like Oconee 3, which had a large circ. flaw at the time
it was discovered, and figure out and back calculate how long each plant had
until they got to that point, but rather just look at the degradation that each
plant has at a point in time or for degradation time at temperature.
So what we have done is recalculated. This information was in MRP-48; it was
just a different column that we had ranked --
MEMBER WALLIS: You mean there are no points where there are no leaks and no
cracks? It doesn't seem to be anything, any data for no leaks and no cracks.
MR. MATHEWS: No leaks or cracks detected in all of these.
MEMBER WALLIS: Oh, there are no leaks and no cracks.
MR. MATHEWS: Yeah.
MEMBER WALLIS: Oh, that wasn't clear to me at all.
MR. MATHEWS: Or cracks.
MEMBER WALLIS: I thought it was that there were no leaks, but there were cracks.
MR. MATHEWS: No, no, no. No leaks or cracks.
The X-axis is now what we're calling equivalent effective degradation years,
which is the same thing that was presented as effective full power years normalized
for 600 degrees Fahrenheit. And I think we even used the term effective degradation
years in the original submittal in MRP-48.
But our ranking system was based on taking each plant's number, at that time,
and then figuring out how many years they had left to be equivalent to Oconee
III.
We said, you know, that's probably not the right way to look at it in the future.
So we are just ranking it. Where does each plant are they? Starting at zero
at zero and going to the highest plant at the time we had the data was Oconee
I, I believe it was.
MEMBER APOSTOLAKIS: So these are the years that are left in the future? No?
MR. MATHEWS: No, no, no. This is accumulated years from time zero to the --
to February 28th. We are going to update all those numbers.
MEMBER BONACA: For understanding, the blue ones, the diamond, no leaks, cracks
detected. Some of them have not been inspected, right?
MR. MATHEWS: No, well, all of the ones that are solid blue have done either
a top-of-the-head visual or a volumetric of their plant.
MEMBER APOSTOLAKIS: So pick a point and explain what it means.
MR. MATHEWS: Okay.
MEMBER APOSTOLAKIS: Let's pick the very first one.
MR. MATHEWS: This point right here?
MEMBER APOSTOLAKIS: Yeah. What does it mean?
MR. MATHEWS: That plant is the lowest ranked unit on time at temperature.
MEMBER APOSTOLAKIS: Okay.
MR. MATHEWS: It is a cold head plant. It's a very cold head plant. And even
though they have been running for a significant number of years, when you normalize
their time at temperature, they are only about one year, effective full-power
year at 600 degrees Fahrenheit.
MEMBER ROSEN: Effective degradation year.
MR. MATHEWS: Yeah. One effective degradation year because they have run at such
cold head temperatures.
You take another plant here --
MEMBER APOSTOLAKIS: Wait, wait, wait. Why is it 69? What does 69 mean?
MR. MATHEWS: There's 69 units and this is just a rank. This is just a sort that
shows the rank.
MEMBER WALLIS: It's not a property. It's just a number assigned to the plant.
MEMBER APOSTOLAKIS: So this is plant number 69?
MR. MATHEWS: It's plant number 69. What it means is that this one has the lowest
time at temperature of all 69 PWRs in the country. This one has the next lowest.
You come on down and they get higher and higher in their effective degradation
years until you get to Oconee I, which had the longest time at temperature run
of all the plants at that time.
MEMBER KRESS: How do you normalize the temperature? Is that linear?
MR. MATHEWS: No, it's an arrhenius equation.
PARTICIPANT: It's a arrhenius equation, okay.
MEMBER KRESS: Is that what accounts for the big split right there or --
MR. MATHEWS: Right. These plants are cold head plants. So when you normalize
it to 600 degrees, they accumulate effective degradation units at a very low
rate in real time. Ones that are over 600, these and Davis-Besse and some of
the others that are slightly over 600 accumulate effective degradation units
at greater than real time.
So, you know, even though they got 21.7 or whatever the number was, their effective
full-power unit was less than that, but they had been running it over 600 degrees.
So to normalize it to 600 would --
CO-CHAIRMAN FORD: But the fact that you have a discontinuity and your algorithm
only takes in temperature, does that give you --
MR. MATHEWS: In the time that you operate.
CO-CHAIRMAN FORD: But the fact that you have a major discontinuity in that relationship
is telling you there is something missing from that algorithm.
PARTICIPANTS: No.
MR. MATHEWS: There will be some plants running cold head temperatures and some
plants run hot head.
CO-CHAIRMAN FORD: But using an arrhenius plot, they should all meld into the
same plot.
MR. MATHEWS: No, no, no. then the other variable is how long that they've been
running.
MEMBER APOSTOLAKIS: But the vertical axis is still not clear to me.
MS. KING: What we did was when we made this calculation for EDY, we just sorted
it from top to bottom and assigned a number one through 69.
MEMBER APOSTOLAKIS: Oh, afterwards you assigned a number? Okay.
MR. MATHEWS: Yeah, we assigned a number after we sorted, ranked on EDY. This
is just the rank of the unit based on EDY.
MEMBER ROSEN: This is too simple for you to understand.
(Laughter.)
MEMBER ROSEN: It's to simple for you to understand. You can't get your guns
down that low.
Now let me go back to my question. The break in the data that I was referring
to was not the one down all the way out in the EDY curve. It's the one up at
five EDY. Do you want to point to that and tell me what that one's about?
MR. MATHEWS: These plants right here are all Westinghouse units that are later
designed and were designed to run with significantly colder heads, somewhere
around T-cold, around 550 to 560 degrees Fahrenheit in the head region. They
have got a lot of bypass flow that goes to the head.
Most of the plants in here were designed with some bypass flow, and you can
call them warm-heads, if you will. They are 580 to 600 degree range.
MEMBER BONACA: The others are hot-heads.
MR. MATHEWS: And these are the hot-head plants --
(Laughter.)
MR. MATHEWS: -- that run at 600 or higher on their temperature on their head.
MEMBER ROSEN: Now some plants have modified that flow scheme during their life.
They have gone from being hot-heads to warm-heads. Some of the warm-heads have
gone to cold.
MR. MATHEWS: Right.
MEMBER ROSEN: Did you take that into account in EDY?
MR. MATHEWS: We took each period of operation at each temperature when we calculated
the effective degradation years, and then we will use their new head temperature
to figure out how fast they move to the right, if you will.
MEMBER KRESS: How did you get the activation years?
MR. MATHEWS: We used -- for this we used 51 kilocalories per mole for crack
initiation.
MEMBER KRESS: Oh, so that's for experiments on crack initiation.
MR. MATHEWS: Yeah. Okay?
MEMBER WALLIS: It has a lot of uncertainty associated with it, I would assume.
MR. MATHEWS: It's not a lot, but there is -- well, there may be. I don't know.
We did some sensitivity studies on our initial ranking going all the way down
to 40 kilocalories per mole to see what impact it had on the stack-up of the
industry and plants moved around a little bit because of different times and
et cetera, but it wasn't a radical shift, and some plants were in a little different
position.
MEMBER ROSEN: Now you acknowledge that this is changing every day, this chart,
right?
MS. KING: Right.
MR. MATHEWS: It should be, but we don't change it every day. In fact, the data
is all effective over a year ago. We are going to update all that data.
MEMBER ROSEN: I understand you wouldn't change it every day, but --
MS. KING: It's expected that the plant would calculate their EDY continuously.
MEMBER ROSEN: Have you done a calculation, a prospective calculation, so that
you know where the plants will end up six months from now, a year from now,
two years from now? Because obviously this picture is changing.
MEMBER KRESS: Other than the temperature problem it just shifts one point.
MR. MATHEWS: Each plant will move to the right at a different speed depending
on what its temperature is. But typically, they are kind of ranked like they
are here. The hot-head plants are here. The cold-head plants are here. And the
warm-head plants are in the middle somewhere.
MEMBER ROSEN: Because each plant moves to the right at a different rate, the
order will change.
MR. MATHEWS: I guess my intent -- and this is my chart. I kind of came up with
it.
(Laughter.)
MR. MATHEWS: --- would be to maintain that initial ranking --
MEMBER ROSEN: Does that mean we can't comment on it?
MR. MATHEWS: Oh, sure, you can.
(Laughter.)
MR. MATHEWS: But, yeah. If I resorted every time I replotted the thing, then,
yeah, the plants would move up and down in the ranking. But probably it would
be more instructive to watch them move to the right at the different paces.
MEMBER ROSEN: I suggest that you press the sort button every once in a while.
MR. MATHEWS: That's probably not a bad idea. Press the sort button every once
in a while.
MEMBER WALLIS: Well, let's tell us the substance now.
MR. MATHEWS: Okay. Now, all of the plants that are red triangles have been inspected
and found leakage.
MEMBER WALLIS: That they've seen deposits?
MR. MATHEWS: Well, yeah. Every one of them has had through wall --
MEMBER WALLIS: They have seen deposits. They have not measured a flow. They
have seen deposits.
MR. MATHEWS: Right.
MEMBER WALLIS: There might have been a leak with no deposit, but the evidence
is the deposit. So those have seen deposits; is that right?
MR. MATHEWS: These two -- well, three plants. We have three plants on here that
are kind of yellow squares. They were plants that did volumetric inspections,
found cracks in some penetrations that were not through walls, but did not have
leakage at that point.
MEMBER WALLIS: And they did not see boron?
MR. MATHEWS: Right, there's no leakage yet.
MEMBER WALLIS: Did not see boron. How do you know there's no leakage?
MR. MATHEWS: Well, they quantify as best they can with NDE at that point in
time the flaws, and the flaws were not through walls, did not reach a pressure
boundary.
There are three of those. This one is the Millstone, and this one was --
MEMBER WALLIS: And there is one that's behind another one.
MEMBER ROSEN: Robinson.
MR. MATHEWS: No, Robinson did a visual and found no leakage.
MEMBER BONACA: A question that I have. For those that were inspected volumetrically
and found cracks, did they fix those cracks? Did they replace the nozzles?
MR. MATHEWS: There's one in here that you can barely see. Cook 2 found a flaw
in '94 and they repaired the flaw in '96.
MEMBER BONACA: Okay.
MR. MATHEWS: Then they came back in 2002, this spring. They did both a visual
and a volumetric on their plant and found no additional flaws anywhere.
Several of these plants, clearly the ones that have yellow have done volumetric
and it's hard. You can't tell from this symbol whether they have done volumetric
or --
MEMBER WALLIS: Yeah, I think that is why I have asked you about it. You said
there are no leaks detected is the main thing. The cracks are somehow inferred
from the leaks in the blues, isn't it?
MR. MATHEWS: Right. The reason it says that is because that triangle encompasses
both visual and volumetric.
MEMBER WALLIS: It would be nice to break that out into two.
MR. MATHEWS: We have a slide and we'll put it up here if she can get to it.
What I have done is flagged the plants that did volumetric in that blue triangle.
MS. KING: It's a little busy, but --
MEMBER WALLIS: Those guys did volumetric?
MR. MATHEWS: All of these plants that have blue have done volumetric in addition
to --
MEMBER WALLIS: So those other blues, say, between 15 and 20, they are just relying
on not seeing "popcorn"?
MR. MATHEWS: Right. These plants have done their 2001-01 response of an effective
visual examination.
MEMBER WALLIS: But we know nothing about the crack situation in those plants?
MR. MATHEWS: Correct. We know they don't have leaks coming to the top of the
head. That's what we know at this time.
CO-CHAIRMAN FORD: But I assume that we are going to discuss that later on when
we come to the whole question of inspection. Maybe it will be in the NRR one,
but this whole question about the relationship between where you see cracks
and where you see "popcorn" or not. That's going to come into--
MR. MATHEWS: I suspect we will get into heavy discussions of that when we talk
about the inspection plan.
CO-CHAIRMAN FORD: Good. While she still has got that slide up --
MEMBER ROSEN: Excuse me. In the blue diamonds, again, it says no leaks, slash,
but cracks were detected. You don't mean that. You mean no leaks or cracks were
detected?
MR. MATHEWS: No leaks or cracks were detected.
MEMBER ROSEN: But if you just pick this piece of paper up, you will get the
opposite piece of information.
MS. KING: We will make sure that gets fixed.
CO-CHAIRMAN FORD: And also you didn't actually know anything about cracks if
you didn't find leaks. So I think you need two different colors, one which is
no leaks detected and another one which has no leaks nor cracks.
MR. MATHEWS: Excel has a limited number of symbols. We are tracking it that
way. We just -- it is kind of hard to get it all on one graph, but I'll try
and do better.
CO-CHAIRMAN FORD: As I mentioned in the very beginning, there was a question
raised about the statistical veracity of this. You could increase that or waylay
that problem by including all the French data, using your algorithm, but on
the French inspection data.
Is that a possibility or do you not even want to approach that?
MR. MATHEWS: Well, I'm not sure we got even as good a handle on French head
temperatures as we have on our own. The other thing is it is not clear to me
that what happened in the French plants is the same thing that is happening
here.
CO-CHAIRMAN FORD: Well, could you expand on that? Because this was the answer
to my question at the very first meeting in July. The French operations got
no bearing at all in the United States operations, and I don't understand that.
Why?
MR. MATHEWS: I think there was significant differences in the processing of
the material that was used.
CO-CHAIRMAN FORD: But processing doesn't come into your algorithm. The only
thing in your algorithm is temperature.
MR. MATHEWS: Time and temperature, you're right. That's right.
MS. KING: But it would affect the inspection results.
CO-CHAIRMAN FORD: Exactly. That's why I am asking why don't you improve the
algorithm. But regardless, if temperature is the only thing in your algorithm,
you should be able to increase your database by including the French data.
MR. MATHEWS: Hopefully, they may all be here and that --
CO-CHAIRMAN FORD: Then that screws up entirely your algorithm.
CO-CHAIRMAN SIEBER: No, it just says there is a difference between the points.
MR. MATHEWS: It says to me that there's something different then --
CO-CHAIRMAN FORD: Their algorithm is not complete, which we know.
MR. MATHEWS: Right. It's just time and temperature. Okay.
CO-CHAIRMAN SIEBER: Well, we know that the heat is apparently very important.
CO-CHAIRMAN FORD: Not in this algorithm.
CO-CHAIRMAN SIEBER: No, but we know it is important to the physical --
CO-CHAIRMAN FORD: Absolutely.
MR. MATHEWS: And what we -- I guess what I hope and what I believe is that the
plants that are out here are the leading edge not only in time at temperature,
but in the bad material, too. And so what we may find -- and personally I expect
to find -- there will be plants that will reach these same time at temperatures
that have no problem.
CO-CHAIRMAN FORD: The reason why I keep hammering on this is that the algorithm
that you've got served a very useful purpose back in July of last year when
you were coming up with your inspection prioritization.
But I hope that it is not the intention of the industry to keep willy-nilly
on this algorithm as if it's the only prediction algorithm in existence because
it is obviously incomplete.
MR. MATHEWS: We know there are other parameters, and when we are able, based
on what we see in the field --
CO-CHAIRMAN FORD: Well, I would hope that from a research point of view it is
not when we are able. I mean, I hope that we have got ongoing work to come up
with this prediction algorithm which we are going to need until all the heads
are replaced. And even then you're going to need it.
MR. MATHEWS: I guess the main problem I see with trying to do it is that all
of the tools that I've seen are based on Alloy 600 base metal, and we've got
several of these plants where the through wall leakage came through the weld
metal.
CO-CHAIRMAN FORD: Well, put yourself in two years' time when I assume that the
staff are going to ask you the question, tell me why my safety posture has changed
significantly; tell me quantitatively why my safety posture has changed by going
to 690 and Alloy 52. Will you be able to answer that question?
MR. MATHEWS: I certainly hope so, and we will be looking into --
CO-CHAIRMAN FORD: Being a researcher, I'm very susceptible to this question
because it takes more than two years to come up with that answer unless you've
already got it in your back pocket.
MR. MATHEWS: Well, I don't have it in my back pocket finally, no.
CO-CHAIRMAN FORD: Okay.
MR. MATHEWS: That was what I had as the introduction, and I'd like to move on
in and get EPRI to come up here and discuss the crack growth rate for Alloy
600 and where we stand on that in the material in the report.
MS. KING: We had this planned as a 45 minute presentation. Do you want to go
into that now or do you want to take a break?
CO-CHAIRMAN FORD: I see. John, does your talk actually go into two parts, fall
into two?
MR. HICKLING: Yes, it does.
CO-CHAIRMAN FORD: Let's take your first part and then we'll break.
MR. HICKLING: Good morning ladies and gentlemen. My name is John Hickling from
EPRI, and I'm going to talk in some detail about a small piece of this jigsaw,
but it is only a small piece, and there are the questions which this presentation
certainly won't answer.
What I'm trying to get to is an agreed crack growth rate for thick section Alloy
600 material exposed to PWR primary water. Everybody knows that Alloy 600 is
susceptible to primary water stress corrosion cracking. We've known that for
a very long time, every since Coriou back in the '60s first discovered the phenomenon.
It's been studied mainly on steam generated tubing where its impact until recently
has definitely been greatest, and the challenge now in terms of head penetrations
is to find out what a thick section material -- how that behaves and to agree
on what sort of crack growth rate we should be using in deterministic and probablistic
analyses.
So the goal here is to establish a generic crack growth rate applicable to this
material, and our approach was to gather together some of the experts in this
field to advise us, and this was done starting in August last year.
Can we flip forward to the slide of the people names? One more.
And we looked around the world for those people who we thought could offer the
best advice on this problem. These are the core team members of the MRP expert
panel.
We've had a lot of people at various meetings. We've had about four or five
meetings of the expert panel since August last year. I myself came into this
field only in December when I joined EPRI, but I have worked on stress corrosion
cracking for very many years.
As your Chairman well knows, it's not necessarily a particularly exact science,
and these are the people who have been in the core team advising us right through.
Can we go back to the overhead?
CO-CHAIRMAN FORD: If I could, just for the other members, apart from Bill, who
don't know these names, these are good people. It's not just a random selection
of experts.
MEMBER KRESS: Are you including this one?
CO-CHAIRMAN FORD: Bill Shack?
MEMBER KRESS: Yeah, called Bill Shack.
CO-CHAIRMAN FORD: He's okay.
MEMBER SHACK: No doubt about one of them.
(Laughter.)
MR. HICKLING: Bill is by definition okay.
MEMBER APOSTOLAKIS: But when you say "expert," you're not conducting
any expert opinion solicitation here, are you?
MR. HICKLING: No.
MEMBER APOSTOLAKIS: It's just that they're advisors to your program.
MR. HICKLING: Not quite. We, as you'll see when I get into the presentation,
we have to look where the data we're using has been generated. So those people
who have generated the data qualify straight away to some extent.
We've also included other people whose expertise is more in analyzing the mechanistic
side of primary water stress corrosion cracking. We've included people whose
expertise is more in analyzing application of data.
MEMBER APOSTOLAKIS: But their role is what? To advise you on the problem.
MR. HICKLING: Their role is to try and reach a maximum degree of consensus on
what the crack growth rates should be that we're using for Alloy 600.
MEMBER APOSTOLAKIS: Okay.
MR. HICKLING: So the work of this expert panel, which started, as I say, in
August last year, falls really into two sections, and that's why I would take
the presentation perhaps in the two sections here.
The first one was to consider following the Oconee experience. What might be
happening in the environment which would exist in the annulus of a crack where
a leak had already occurred, i.e., we're talking about external OD cracking
in that case.
And I'm going to take that issue first in this presentation and then come back
to the rather large body of work which is on the actual crack growth rate under
normal PWSCC conditions.
I put in a little bullet here and will come back to that right at the end of
the presentation about Davis-Besse. The expert panel or a subgroup of it met
quite recently to consider the implications of the Davis-Besse incident to this
argument and has reached the conclusion that the arguments I'm presenting today
are basically valid in a non-Davis-Besse situation, i.e., at low leakage rates.
And I have a couple of comments to make about how we think the Davis-Besse environment
might affect that growth rate.
Next one, please.
So if we move through the presentation on to how we are trying to use it, I
think we'll go straight on to the external OD environment.
Slide. Thank you.
A lot of thinking was put into this, first of all, as to what the most and likely
environment would be once you had a through wall crack in a CIDM nozzle, and
the conclusion was there were three likely environments, and they depend to
some extent on the situation as the leak develops because intragranular stress
corrosion cracking, primary water stress corrosion cracking in Alloy 600 leads
to extremely tight, highly branched cracks.
So that the first time that a crack penetrates the material, the OD surface,
the leakage rate is likely to be extremely low, and the pressure drop is likely
to be taking place purely within the crack.
So the environment at that stage is almost certainly going to be hydrogenated,
super heated steam.
MEMBER WALLIS: So where does the boron go? If you've got boron coming in with
the water, it can't just turn to steam. The boron has got to go somewhere.
MR. HICKLING: No, the boron will exit also with the steam.
MEMBER WALLIS: Well, it so. So it's steam carrying boron in some form.
MR. HICKLING: Yes, yes.
MEMBER WALLIS: So it's borated, super heated.
MR. HICKLING: Borated, super heated state.
MEMBER KRESS: That depends on the pressure at which you convert it into steam.
MEMBER WALLIS: Just by continuity.
MEMBER KRESS: If the pressure is very high, it will concentrate in the water.
If the pressure is very low, it's going to go out with the steam. So I don't
know how you --
MEMBER WALLIS: Continuity has got to go out some --
MR. HICKLING: It depends on leakage path and the hydraulics of the situation.
MEMBER KRESS: That's what I'm trying to say, yeah.
MR. HICKLING: Absolutely.
MEMBER ROSEN: Now, the boron in the water will range from, depending on the
cycle, from something like 2,000 parts per million down to very low, maybe 100
parts per million.
It will also characterize the boron in the super heated steam, or is there a
partition factor?
MR. HICKLING: I think that's not an issue in this case for the super heated
steam environment. If you see, looking down the slide, we have the three environments.
We have the two extreme cases, at the beginning, when we're dealing almost certainly
with only steam in the annulus, and we have the second case where we've already
flooded the annulus, much later where we have a very high leak rate.
I'm not saying we've got wastage or corrosion or cavity formation. I'm saying
we have flooded the annulus with liquid so that the boiling point is high up
in the annulus, well above the J-grove weld.
And remember that the J-grove weld determines where the cracking is going to
occur because of the residual stress consideration.
And then the third point, which is what we've attached most attention to, is
what would happen if you're getting considerably boiling and partition at the
point of exit from the crack, i.e., you're getting a different environment forming
exactly at the point where you have your residual stress, and that is what most
effort has bene put into.
MEMBER ROSEN: Well, are you implying that the concentration of boric acid to
be higher than the concentration in the primary water?
MR. HICKLING: Yes. Oh, yes.
MEMBER ROSEN: The concentrates?
MR. HICKLING: Oh, yeah, and the lithium hydroxide does, too.
MEMBER ROSEN: Ultimately it concentrates, but at the very first instance, I
guess it's not that relevant. At the very first instance, there's a little boron.
Perhaps what the partition factor between steam and water doesn't really matter
as long as a little carryover.
The water that carried over stays there.
MR. HICKLING: Yeah.
MEMBER ROSEN: And then it continues to build and build.
MR. HICKLING: Yeah. The steam environment, because it's a pure super heated
steam environment with the exception of the boron and lithium carryover, is
basically not a difficult environment to handle because there's been a lot of
work done on that. The --
MEMBER WALLIS: I'm wondering about that. I mean it depends on where boron and
lithium goes. If it builds up, if it deposits on the walls, then your environment
is essentially walls plated with boron in various --
MR. HICKLING: Are you talking about the walls of the crack or the annulus?
MEMBER WALLIS: Of wherever the steam is coming out and impinges upon. The OD
annulus environment here.
MR. HICKLING: Yes.
MEMBER WALLIS: And presumably some boron is carried out by the steam, but it's
a very low flow rate. It's a big area in that.
MR. HICKLING: Yes.
MEMBER WALLIS: I would think it would fill up with boron crystals or whatever,
the popcorn or whatever.
MR. HICKLING: Yes, very good point.
MEMBER WALLIS: So the environment, what the wall sees is whatever the bottom
of those crystals' condition is, which presumably is dry or wet or whatever,
depending on the various phases of boron, boric acid with temperature and concentration.
MR. HICKLING: Correct.
MEMBER WALLIS: So it could be doing something to the wall because it's concentrated
boric acid. It's not steam that the wall sees.
MR. HICKLING: Yes. You'll see it right at the end when I come back to talk about
the Davis-Besse situation. There's a little -- we have very, very little data
on stress corrosion cracking of Inconel in concentrated boric acid solutions.
There is one paper essentially resulting from one French program which has addressed
that particular condition.
The main concern behind the consideration of the environment in this case on
the OD environment has always been traditionally caustic and caustic formation.
MEMBER WALLIS: That puzzled me. That's what Bill was telling us earlier. I guess
he can't tell us anything now.
How does it get to be caustic when there's so much boric acid there?
MR. HICKLING: Because the concentration mechanism that is taking place here,
depending upon the interactions and particularly the precipitation, as you correctly
pointed out, you are going to get precipitation and plugging, and depending
upon the exact way in which that forms, you can postulate different chemical
environments which might form.
And you cannot per se rule out the tendency to go caustic, and as was also mentioned,
you have to consider the differences in boron concentration between beginning
and end of cycle, which will affect potentially the final pH of that concentrated
solution, and all of that was taken into account.
MEMBER ROSEN: And the fact that there's a coordinated lithium being used in
many plants.
MR. HICKLING: Absolutely.
MEMBER ROSEN: The pH of the rapid coolant during normal operation is typically
not above neutral. It is basic, kept in the 7.0 to 7.4 range, I would guess.
MR. HICKLING: Right, yes.
MEMBER ROSEN: Now, that does not characterize the pH in the crack.
MR. HICKLING: The pH in the annulus, if you're having boiling in a concentrated
environment.
MEMBER ROSEN: Will drive it acidic?
MR. HICKLING: I'll get to that in two minutes, if I may. Let me take the two
simpler environments first because the simpler environment -- well, no, I'm
sorry. One more slide, Christine, please.
There's one consideration I'd like to take first of all before considering the
three environments because it's a very important one, but it is actually the
same arguments apply to all three potential environments, and that is the extent
to which you might get an oxygenated condition developing within the annulus
low down, just above the J-groove weld where you're expecting a stress corrosion
cracking to occur.
And traditionally, of course, oxygen virates' (phonetic) effect on electrochemical
potential has a huge potential impact on cracking susceptibility. So the panel
spent quite some time looking at the arguments as to whether or not the crevice,
right down in the crevice, could be oxygenated.
And there are various ways that that was done. The first was to use some back
diffusion models for oxygen. In fact, two independent assessments were made.
Considerations of oxygen consumption along the metal walls --
MEMBER WALLIS: But does it just diffuse? I mean, there's a flow pattern in this
annulus.
MR. HICKLING: Yeah.
MEMBER WALLIS: There's a crack at one place producing a jet of some sort. I
would think it's not just diffusion that's going on. You have to analyze the
fluid flow pattern in that space.
MR. HICKLING: Correct.
MEMBER WALLIS: There's a mechanism for back flow in the place where the jet
is not perhaps.
MEMBER ROSEN: In fact, the jet could be pumping the crack, right?
MEMBER WALLIS: But I don't know if it can. We'd have to see an analysis.
MEMBER ROSEN: Like a jet pump in a BWR, just like a jet pump.
MR. HICKLING: You've got to remember that we're talking here about a very, very
narrow, deep annulus --
MEMBER ROSEN: Around the grain.
MR. HICKLING: -- at this point.
MEMBER WALLIS: But again, I haven't seen any equations or figures or anything.
MR. HICKLING: Right, yes.
MEMBER WALLIS: So I'd have to look at the model to see whether -- when you say
"diffusion," it makes me a little suspicious. If someone assumed it
was diffusion, I doubt if that's what was going on.
MR. HICKLING: No, the model, both of the model concerned, in fact, do take that
into account. I think probably more important in concluding that oxygen is not
present right down at the bottom of this very deep and narrow crack, also some
of the other points, the oxygen consumption, the presence of hydrogen itself
because, of course, hydrogen is present in the water and by diffusion through
the metal of the head and is available to react with any oxygen that might be
there.
And finally, the fact that even if you were to postulate very low oxygen levels
still being credible at the bottom of the crevice, you do have a coupling effect
between the alloy steel and the Alloy 600, a galvanic coupling effect, all of
which will keep the potential low.
MEMBER WALLIS: Why does the hydrogen react with the oxygen here when it doesn't
in the containment?
MR. HICKLING: This isn't --
MEMBER WALLIS: After putting miters (phonetic) in there?
MR. HICKLING: We're talking about reaction here within an aqueous phase.
MEMBER WALLIS: Oh, okay. So that's much more graphic.
MR. HICKLING: Yes. So the bottom line conclusion of all of these considerations
was that it is not necessary to consider an oxygenated crevice condition right
down at the bottom. As I said, this analysis does not treat a wastage in cavity
formation situation.
MEMBER WALLIS: Is there any real evidence of non-oxidation in this annulus space,
observation of no rust?
MR. HICKLING: I think the answer to that has to be that there is no observation
of what that crevice looks like right down at the bottom.
MEMBER WALLIS: That would have been destructive examinations of real cracked
--
MR. HICKLING: The only one I'm aware of is in Bouget nozzle that first cracked,
which was destructively examined, in fact, and there was no real evidence of
--
MEMBER WALLIS: Thank you.
So the fact is it was useful.
MR. HICKLING: Oh, yes, yes.
CO-CHAIRMAN SIEBER: If wastage does occur, then these arguments, except for
corrosion potential, then fall apart; is that correct?
MR. HICKLING: I'm sorry. I didn't hear the first part of the question.
CO-CHAIRMAN SIEBER: If wastage does occur--
MR. HICKLING: Yes.
CO-CHAIRMAN SIEBER: -- okay, then these arguments about oxygenation fall apart
because the geometry is now changed.
MR. HICKLING: If significant --
CO-CHAIRMAN SIEBER: With the exception of corrosion potential; is that correct?
MR. HICKLING: Correct. If significant wastage and cavity formation were to occur,
then this is a different situation, which would require separate consideration.
CO-CHAIRMAN SIEBER: Now, aside from the factor of the wastage weakening the
basic structure of the head, the added oxygen would increase the crack growth
rate significantly, don't you think?
MR. HICKLING: Not necessarily. Primary water stress corrosion cracking of Alloy
600, Alloy 600 has a number of separate modes of stress corrosion cracking,
and your conclusion would be correct for some of them, but not to primary water
stress corrosion cracking.
Remember the original finding that Alloy 600 cracks in pure water or in PWR
primary water is extremely surprising, and the mechanistic reasons for it doing
that are very closely linked with the fact that the electrochemical potential
--
CO-CHAIRMAN SIEBER: Is there.
MR. HICKLING: -- is established in the region of the nickel/nickel oxide transition.
CO-CHAIRMAN SIEBER: Right.
MR. HICKLING: And that is a low potential phenomenon. So in that case it's not
fair to assume automatically that oxygen would be negative. It was just a consideration
that needed to be very carefully looked at in terms of the narrow annulus.
Now, I'll come back to make a comment right at the --
MEMBER SHACK: Because it could be cracked by another mechanism, and you would
have to address that one.
MR. HICKLING: Absolutely, yes. If you had an oxygenated environment, a highly
alkaline environment, then that is not primary water stress corrosion cracking.
It's a different mode, I think.
MEMBER WALLIS: Tell me more about the hydrogen. I mean, we were hearing about
hydrogen explosions in BWRs where they had essentially a stoichiometric mixture
of hydrogen and oxygen resulting from radiolysis (phonetic).
MR. HICKLING: Yes.
MEMBER WALLIS: So there is an oxygen in there, not just all leaking out necessarily
by the hydrogen.
MR. HICKLING: In the PWR, primary water environment, that is your main reason
for adding large over pressures of hydrogen, to make sure it is all --
MEMBER WALLIS: So these are all hydrogenated plants?
CO-CHAIRMAN SIEBER: Yes.
MR. HICKLING: All PWRs run with high hydrogen levels for that reason.
So that consideration was the elimination of oxygen from the picture for the
narrow crevice at the beginning of the situation, the non-wasted situation.
Looking back then at the three environments that were considered, and the first
one is hydrogenated steam, and as I mentioned, there is quite a lot of evidence,
quite a lot of information available on the way in which Alloy 600 cracks in
hydrogenated steam primarily because hydrogenated steam has been used as an
accelerated test method for determining crack susceptibility in this and other
nickel based alloys.
And the main conclusion of the data that's available is that in terms of pure
hydrogenated steam, and not including boron or lithium in this, the impure steam
environment which is used to accelerate cracking involves chloride and sulfate
as contaminates.
In terms of the hydrogenated steel environment which you would expect at the
beginning with a very tight crack, the rates of cracking are going to be virtually
the same as they would be in normal primary water at the same temperature.
MEMBER WALLIS: Is cumulative percent with IGS? So that means that after 1,000
hours, 60 percent of them have cracks?
MR. HICKLING: Yeah. This is one diagram picked out of a -- it's very hard to
summarize in some cases all of the work that's been done on Alloy 600. This
particular issue has been studied for very many years, particularly at the Westinghouse
laboratories from about 1987 through '95.
MEMBER WALLIS: My question really was this crack development is so rapid because
the temperature is so high. Isn't that why?
MR. HICKLING: Correct.
MEMBER WALLIS: If we looked at this as typical, we'd be really scared.
MR. HICKLING: yes.
MEMBER ROSEN: You see, now that's the danger of coming to ACRS. We start putting
things together.
If you just said that these crack rate growth rates are accelerated tremendously
in chloride and sulfate environments, chloride and sulfate contamination --
MR. HICKLING: Contamination, yes.
MEMBER ROSEN: -- did that happen at Davis-Besse?
MR. HICKLING: I'm going to deal with what we know about that, and I know nothing
whatsoever about the Davis-Besse situation. I have no reason to believe it did,
but I'm going to --
MEMBER ROSEN: That's a question we could perhaps ask the staff with the applicant.
It's easy to get chloride contamination in the primary from a leak from the
secondary side. If your secondary side has a, you know, brackish or that kind
of water, you're going to be -- in your cooling water, you're going to have
chloride.
So if you get some sort of ingress into the secondary side, you will have chloride
contamination in the secondary side. It's possible, although not likely to have
an intrusion into the primary system.
CO-CHAIRMAN SIEBER: I'm not sure how that happens since the primary runs at
a higher pressure.
MEMBER ROSEN: Yeah. That's why it's difficult, but it can happen during shutdown
or --
CO-CHAIRMAN SIEBER: It's like pushing water uphill.
MEMBER ROSEN: Well, yes, but it's not always true that the primary is higher
than the secondary. You can have chloride contamination in the primary or sulfate
contamination.
CO-CHAIRMAN SIEBER: I'd have to think about that. It doesn't pop to mind readily.
MEMBER ROSEN: No, I'm talking about in shutdown modes.
CO-CHAIRMAN SIEBER: Oh, all right.
MR. HICKLING: Let me just point out that when I said impure steam as a test
environment, I'm talking about very considerable levels of chloride contamination,
much larger than you could ever postulate, I think, in terms of an accidental
contamination of the primary circuit.
MEMBER WALLIS: Unless it concentrates in some way.
MR. HICKLING: Correct, but this was referring to the hydrogenated steam environment.
MEMBER WALLIS: But it came in as water and is going to go back again. And so
did it get carried out with the steam or not?
MR. HICKLING: The second OD annulus environment which we're going to talk about
in detail in terms of likely crack growth rates that have to be assumed is then
normal primary water, which could definitely be the case once the annulus is
flooded and when boiling is not taking place down at the bottom of the annulus
where you might be expecting OD cracking.
MEMBER WALLIS: So someone has worked all of that out in terms of heat transfer
rate? Because with the hot head you would expect that it would boil or flash
pretty quickly, wouldn't it?
MR. HICKLING: Yeah, well, in terms of boiling or flashing, they're all going
to flash quickly. The head temperature differences are minor in terms of the
phase changes which go on.
MEMBER WALLIS: Right.
MR. HICKLING: Although they do have cracks.
MEMBER WALLIS: Doesn't it take a pretty big leak to get any boiling at all in
the annulus?
MR. HICKLING: Well, it will take a significant amount of leakage before that
scenario takes place, yeah.
So the environment which attracted most attention in terms of the expert panel
is the environment number three of the concentrated PWR primary waters as a
result of boiling, and the caveat on this is that these considerations apply
to low leak rates, and the panel has adopted a definition of less than one liter
per hour to quantify what we're talking about here, which is pretty low leakage,
in some cases very much less.
There are various ways in which we can analyze the problem of what environment
is formed and particularly whether or not caustic forms and pH. One of them
is to use the thermodynamic calculations, which are available, which have been
produced largely because of secondary side stress corrosion cracking in steam
generators, a phenomenon which has been studied very, very intensely over many
years.
And EPRI has a program called MULTEQ, which will calculate the expected pH as
you concentrate up an environment of that sort, and the answer that comes out
by using that program is that you would expect a high temperature pH of somewhere
initially between 4.0 and 9.4. So it's quite a narrow range that, in fact, due
to the composition of the liquid which is being concentrates.
In fact, that pH range is probably far too broad as calculated because as was
correctly pointed out, you're going to get precipitation of various insoluble
compounds. We know that, and that narrows it down because it has a buffering
effect.
So the likely pH range is going to be much smaller than that. What experimental
evidence do we have for what pHs might be involved? After the Bouget experience,
the French -- next slide, please -- did a very interesting experiment. This
is CEA, the French atomic laboratory, which simulated leakage in this case by
injecting the liquid, which was to be concentrated through a heated block, blocking
off the flow of liquid so that when it exited the nozzle, there was a very,
very tight leak path exiting, simulating what might be expected from a strained
granular stress corrosion crack, and allowing that vapor to impact on a heated
plate of low alloy steel material simulating the vessel head.
And the next picture gives some feel for what actually happens. In fact, you
do get a huge amount of precipitation occurring in the annulus.
Now, there was one caveat unfortunately on this experiment, which was the --
there was a considerably amount of cooling generated of the low alloy steel,
relevant certainly to the Davis-Besse incident, but not relevant perhaps to
the conditions initially in an annulus where the leak rates are very low and
where you would not expect local cooling of the head.
MEMBER ROSEN: Is that some red rust I see there?
MR. HICKLING: Yes. That is the low alloy plate which is being corroded both
by boric acid corrosion and impingement and simply, you know, moist atmosphere.
So it is rusting.
Okay. We go back to the previous slide.
There's been a second experiment, again, performed in France to look at this
particular issue, and the results of that were published only very recently,
in fact, a month ago. And this involved a slow concentration of a fixed volume
of primary water in an autoclave system, which they considered realistic to
simulate what would be happening.
And the interesting factor here, in fact, after a concentration factor of 1,000
was that the pH was acid, slightly acid, 4.5 rather than alkaline.
So the general conclusion from both the theoretical analysis and the experiments
we know about is that the caustic formation can almost certainly be ruled out.
The pH is going to be very limited. It's certainly not going to move strongly
alkaline. If anything, it's probably likely to move slightly acid in that environment.
MEMBER SHACK: In that French test, I mean, that was done in what? Did the autoclave
have nickel and a low alloy steel?
MR. HICKLING: Yeah. They, in fact, set up a whole system called EVA, and I've
forgotten what EVA stands for, but it was for a simulation of what would be
expected to happen as you concentrated a limited volume in contact with low
alloy steel and nickel, and it was not a simple autoclave, cook-it-up test at
all. It was a leak and bleed and reconcentrate test involving quite a complicated
experimental system. It was published in Avignon, the Avignon conference last
month, yes.
Then the issue came up earlier in connection with the steam: can we exclude
the possibility of contaminants which are known to promote more rapid cracking
of Alloy 600, and in particular, chloride and sulfate, which might be involved.
And it's very difficult to make any absolute sweeping, generic conclusions here.
Obviously the practice during assembly of the heads was to clean. So the amount
of contamination that would have been left after assembly is expected to be
relatively low.
And, secondly, we know that there's going to be considerable steam flushing
within the annulus, which would help to drive out any initial deposited contamination
from assembly of the head.
The expert panel did some calculations of possible concentrations, maximum concentrations
that could ever be expected, even making very negative assumptions as to contamination
which might have been encountered during fabrication, and they were orders of
magnitude below the levels at which you would expect any effects on primary
water stress corrosion cracking.
MEMBER ROSEN: But is that the only way they thought about getting chloride and
sulfite into that crack? Did they think about it as a contamination event of
the primary coolant system and then the chloride and sulfates exiting with the
steam?
MR. HICKLING: Not specifically because I think there's some -- as the discussion
earlier showed, there's some doubt as to whether that is a significant possibility
that you could have a contamination of the primary system by chloride and sulfate
in the way that you could get --
MEMBER ROSEN: Well, if someone were to inject chloride, for example, into the
primary system?
MR. HICKLING: Well, I think that's something that we would very much hope the
water chemistry monitoring and guidelines would prevent.
MEMBER ROSEN: It wouldn't be intentional. Let me say that.
MR. HICKLING: Yes.
MEMBER ROSEN: But it has happened.
MR. HICKLING: The more likely scenario is resin intrusion, and that has been
considered.
CO-CHAIRMAN SIEBER: That's the only place I --
MEMBER ROSEN: Resin from the?
CO-CHAIRMAN SIEBER: Let-down system, yes.
MEMBER ROSEN: That's happened, too. So there are several mechanisms I can point
to.
MR. HICKLING: Yeah, but you've got a huge volume of water in the primary system
to dilute that.
CO-CHAIRMAN SIEBER: And those instances are rare and easily detected.
MR. HICKLING: Yes.
MEMBER ROSEN: But I'm only asking the question, Jack if that happened at Davis-Besse
because it has happened elsewhere. Two bulk mechanisms: injection when the chemists
were trying to -- through they were injecting something and they were actually
injecting something else, and resin releases from the clean-up system.
And I don't know the answer to that question, whether there is any evidence
that happened at Davis-Besse, but I know it has happened elsewhere.
MR. HICKLING: The bottom line of the panel's consideration on the OD annulus
environment was that even in concentrated PWR primary water, we're considering
a very narrow pH range -- there's a typo which is entirely my fault on this
first slide. It should read between 5.0 and 7.5.
And even if we take a pessimistic view and rule out what we k now about precipitation
of buffering so that we're looking at a whole range between about five and nine,
there is only a very, very slight effect on crack growth rate of changes in
pH in this area.
If we just flip forward, please to the next slide, the data in this area was
generated mainly at Ohio State University on Alloy 600 specimens from steam
generator tubing, but there's no reason to believe that in terms of pH effect
that it should be invalid or have any less relevance to what we're considering
here.
There three diagrams are showing between a pH of five and nine the effect on
intragranular stress corrosion crack growth rate at three different stress intensities,
20, 40 and I believe that's 60 at the bottom.
MEMBER WALLIS: That's a freak point, that first graph?
MR. HICKLING: No. If you look at the Y axis on the first diagram, you'll find
it's expanded relative to the other two. You've only got one order of magnitude
difference here, whereas these two are showing two orders of magnitude.
There's no doubt there is a turn-up after about 7.5 pH, and this is to be expected
because if you go sufficiently caustic, then you will get a very rapid increase
in crack growth rate.
MEMBER WALLIS: It only really occurs in that top figure. It's very different.
MEMBER BONACA: No, no, because --
MEMBER WALLIS: Yeah, but then it comes back down again.
MEMBER BONACA: It's like the midpoint in your other figures.
MEMBER WALLIS: It's like the midpoint in the other figures, but then there's
a point later on above -- I count above nine there, which comes back down again.
So I don't know if it's a real turn-up or not.
MR. HICKLING: Yes. You've got to remember that it's very difficult in testing
at low stress intensity to get a uniform, reproducible crack growth rate anyway.
My inclination is much greater reliance on the low occurrence where there are,
in fact, far more points.
But the bottom line, if we go back, is still the expert panel considered taking
even this extreme pH range you would not expect more than about a factor of
1.5 or 1.6 on crack growth rate over that pH range.
And the recommendation was that within the high temperature range of four to
nine, we should apply a factor of two on whatever crack growth we were proposing
in normal primary water to cover possible uncertainties in the environment.
MEMBER WALLIS: And that crack growth rate is uncertain by more than factor of
two anyway.
MR. HICKLING: That's the second part of the talk, yeah, and we'll get into that.
I guess you may want to take --
CO-CHAIRMAN FORD: John, I can follow your argument, and it's fairly clear. However,
on this particular rationale, you're honing in on crack growth rate. How about
crack initiation, and especially crack initiation density? Because that would
have an effect on the safety analysis.
MR. HICKLING: Yes.
CO-CHAIRMAN FORD: You're propagating for a circumferential crack all the way
around the tube. Are there any comparable data for the effect of the environment
change on crack initiation density?
MR. HICKLING: I'm not immediately aware of data in that pH range on crack initiation.
I don't think it's necessarily relevant to what we're trying to achieve here
though, Peter, because we're trying to disposition here flaws, and as we'll
see in the second part of the discussion, we're trying to disposition flaws
which are already of considerable size.
There's a whole lot of issues about initiation in Alloy 600 which we're jumping
over in this analysis quite deliberately because we're postulating that we already
have relatively deep flaws in order to make the analysis.
CO-CHAIRMAN FORD: From one point, not all the way around, not a 360 degree crack.
MR. HICKLING: Again, when we come onto the way we intend to use what we're proposing,
you'll see that we're not proposing to disposition OD flaws. We'll come on to
see that we're talking about hypothetical arguments about how quickly they could
grow.
CO-CHAIRMAN FORD: Okay.
MR. HICKLING: But we considered the only way to handle the OD crack growth rate
is a probablistic one.
CO-CHAIRMAN FORD: Okay.
MR. HICKLING: Interest in developing a consensus crack growth rate in normal
primary water is in terms of ID flaws how we get to the first leakage rather
than in terms of OD flaws.
CO-CHAIRMAN FORD: Okay. I've got one other question. Sorry.
Apart from the one french data where they measured pH rather than inferred it,
that's the only experimental data of what that annulus environment would be
in terms of pH. Are there any experiments planned or ongoing to increase the
database with specific reference to the effect of leak rate?
MR. HICKLING: Yeah. Firstly, it's not the only experiment. It's the only experiment
-- you're quite correct -- where they specifically measured the pH of the environment.
But the experiments, and Glenn White will be talking about this in addressing
the wastage issue later, there have been experiments performed in this country
as well with two prototypical mock-ups in terms of generating wastage in an
annulus, and although the pH was not measured directly as far as I'm aware in
either of those experiments, the results in terms of wastage of low alloy steel
quite clearly show that, if anything, there's a strong move in the acid direction
once leak rates become very high, much higher than what we're considering here.
CO-CHAIRMAN FORD: Okay.
MEMBER WALLIS: You have this multi-calculation.
MR. HICKLING: Ye.
MEMBER WALLIS: OS pH is four. Now, if you had suitable deposits in that annulus
which you could postulate, you could achieve a much lower pH, couldn't you?
In other words, is there some limit to the pH achievable with --
MR. HICKLING: I think, again, it's a key question of the amount of leakage and
the assumptions you make. I think it's conceivable that you can certainly go
lower. The buffering is preventing you getting to a caustic condition, which
remember was the original consideration.
MEMBER WALLIS: Yeah, but we don't have much of a database. We have some theoretical
calculations. We don't know much about what's really going on in there, and
if you looked at some extreme scenario in which you built up deposits, you could
tell us what the pH could be in the worst case.
MR. HICKLING: Well, as you'll see when we go on to discuss in detail the thermal
hydraulic analysis of the wastage situation, I think you can postulate certain
cases where you might go very acid, yes.
MEMBER WALLIS: I thought so, too, but I haven't seen any figure yet. So I have
to imagine what might be going on in there.
MR. HICKLING: Right.
MEMBER WALLIS: And I can conceive of a scenario where you could have a very
low pH.
MR. HICKLING: I think that's quite correct, but just jumping ahead, it's a point
I was going to make right at the end. Alloy 600, the original design basis for
choosing that material was its resistance to cracking in acid solution. And
so there's no reason, even if you went very acid, to assume that that would
automatically be negative as regards the --
MEMBER WALLIS: So it's a bounding pH rather than a calculated pH.
MR. HICKLING: Yes.
MEMBER WALLIS: Yeah.
MR. HICKLING: That's the natural break because we now go on to the crack growth
rate database.
CO-CHAIRMAN FORD: Thanks a lot, John.
I had a question for you, Larry, which is more of an administrative question.
I notice John has got a few more slides, and I suspect there will be some questions.
I'm proposing that we stop until five minutes to 11, but I notice that Glenn
needs an hour for his presentation. So I leave it up to you and John to work
out how you want to do --
MR. MATHEWS: And then we have Pete's presentation also.
CO-CHAIRMAN FORD: Pardon?
MR. MATHEWS: We have Pete Riccardella's presentation also. So we're running
quite a bit behind here.
CO-CHAIRMAN FORD: Yeah. The trouble is we want to hear them all.
MR. MATHEWS: We can be here all day.
MS. KING: Why don't we come back with a proposal?
CO-CHAIRMAN FORD: Okay, fine. Let's stop until five minutes to 11. Let's go
into recess until then.
(Whereupon, the foregoing matter went off the record at 10:42 a.m. and went
back on the record at 10:55 a.m.)
CO-CHAIRMAN FORD: Okay. We're back in session.
Christine.
MS. KING: Okay. What we would propose --
CO-CHAIRMAN FORD: Yes.
MS. KING: -- since we have a lot of interest in Davis-Besse type issues, we
would like to propose to bring Glenn White's presentation forward --
CO-CHAIRMAN FORD: Good.
MS. KING: -- to this morning following John.
CO-CHAIRMAN FORD: All right.
MS. KING: And depending upon where we land around lunch, I guess we'll either
take lunch or continue into the PFM, and it shouldn't put us too far off schedule
because there was 45 minutes set aside in the afternoon for Glenn.
CO-CHAIRMAN FORD: Okay. So I think what we'll do is we'll have John -- finish
off John.
(Laughter.)
CO-CHAIRMAN FORD: I didn't mean that literally.
And then we'll have Glenn, and then we'll take just three quarters of an hour
lunch, and then we'll catch up time that way.
John.
MS. KING: Okay.
MR. HICKLING: The second part of this presentation deals with the meat of the
work of the expert panel over the last six to eight months, which is what would
be a representative crack growth rate for Alloy 600 base material, thick spectrum
material.
And the initial approach taken was to look at what we've learned in stress corrosion
cracking testing over the last five to ten years particularly where the international
community has focused very much on issues of data quality because it doesn't
matter how sophisticated your statistics or your analysis is later. If your
data is bad quality, it doesn't really allow you to get a handle on stress corrosion
cracking.
And the first thing the expert panel did was to make a list and discuss in depth
some of the key technical issues on crack growth rate testing which need to
be addressed and which conform the basis of screening out suitable higher quality
data from data which is of lower quality for the purpose we are using it.
Many investigations in this area have been for different purposes, trying to
understand the mechanisms, trying to understand effects of off-chemistry, things
like that, and we were trying to get to where we could screen out things like
that.
And as you see, there's a whole list of factors here which involve chemical
environment, loading, the way the material was used, the sort of specimens which
were generated, the loading characteristics during the test, the crack growth
rate monitoring, and all of this sort of thing.
How did we actually do the screening? Really it involved three iterative steps.
The first step was to go back to the laboratories which had generated all the
data we were able to collect worldwide on thick section Alloy 600 material and
ask the initiating laboratory to reexamine their own data in the light of these
criteria we had put up and in the light of discussions which they had been involved
in on the expert panel.
And this probably was the most important step because it led to elimination
of a lot of data points by the initiating laboratory who declared these points
to be unsuitable for this particular purpose for developing a crack growth rate
disposition curve.
MEMBER WALLIS: It didn't eliminate anything because they didn't want to believe
it.
MR. HICKLING: I would hope not. I think the people concerned, their integrity
was such that would not be the case.
The second step was a screening step which EPRI put in place, and it basically
covered two main areas. As I say, we involved international laboratories. You'll
see the list of laboratories in a second, and in one or two cases we had some
difficulty in direct contact with laboratories concerned.
One of them particularly, one European laboratory, had performed tests where
they had only ever reported maximum crack growth rates during the test. Since
the whole thrust of the analysis is to use average crack growth rates determined
in a particular specimen, we could not use that particular data.
So in the end, after trying to obtain , and we put a lot of effort into it,
we had to screen out that particular laboratory's data.
The second point was one which I was very concerned about. We've mixed in this
database specimens which are actively and passively loaded, i.e., they're all
fracture mechanic specimens of one type or another, but some are actually tested
in a tensile testing machine under active load, and some were under displacement
loading, usually by means of wedges.
And the stress corrosion cracking community has known for a number of years
that these are actually or even though you are nominally at the same K value,
you can get a difference in response. It's much more difficult to initiate crack
growth from a passively loaded displacement controlled specimen uniformly.
And so we went back and reexamined the data from that type of specimen, the
wedge open loaded specimens and eliminated all of those specimens where crack
growth had been very non-uniform, and the criterion we used was less than 50
percent initiation across the width of the specimen. And what that does is it
eliminates lots of artificially low points.
Finally, the third iterative step in the screening was for the whole expert
panel to reexamine the borderline cases and what we had done to the database,
and that was done at the beginning of March.
We didn't even bother to start to try and consider numerous tests where no stress
corrosion crack growth was actually obtained, a zero result, and the reason
for that is there are a whole number of reasons why you may get a zero result.
You may have a very non-susceptible heater material, but you may also have done
the test in an inappropriate way. So there's no zero crack growth rate data
in this database at all.
MEMBER WALLIS: That's a bit strange. I mean, in terms of the probability of
a crack occurring, zero cracking would be a good data point, wouldn't it?
MR. HICKLING: In some ways, yes. It does hurt to have to eliminate those points.
That's correct. But in terms of trying to get at crack growth rates, unless
you can convince yourself that everything else was perfect, and it's very difficult
to do, you just have to take that step.
MEMBER WALLIS: You're not interested in initiation.
MR. HICKLING: Correct.
MEMBER WALLIS: You're just interested in growth rate.
MR. HICKLING: Yes, absolutely.
The result of this screening was that we eliminated no less than 203 crack growth
rate data points for one or more reasons, and these reasons are documented.
The main reason is individually documented in the report the MRP is in the process
of issuing on this exercise.
The consolidated database now contains 158 points for average crack growth rate
during each test, and this is consistent basically with the ASTM recommended
procedures for measuring fatigue crack growth rates, to use the average, and
they're plotted at a single representative K value for the data point concerned.
And there, again, there was a certain amount of judgment sometimes involved.
The expert panel was involved in that in detail because the K value in some
tests will change during the test, and we satisfied ourselves that we had a
representative value.
MEMBER LEITCH: Why would you not consider-- several bullets back --
MR. HICKLING: Yes.
MEMBER LEITCH: You mentioned that there was some data that you discarded, eliminated
from consideration because the experiment only considered maximum --
MR. HICKLING: Yes.
MEMBER LEITCH: -- crack growth data. Why would you eliminate that data? Would
that not be the conservative thing to include that data?
MR. HICKLING: Only at first glance. The problem there is that we had no detailed
-- I'm sorry. Let me back up one stage.
The way these tests are run is to use an air fatigue pre-crack in usually a
compact tension specimen, sometimes a DCB specimen, which produces a transgranular
fatigue pre-crack. You then have to go through a second stage in the text where
you initiate an intragranular stress corrosion crack from that transgranular
fatigue pre-crack.
And one of the key things we insisted on was we had to have fractographic information
available on each specimen or at least in the form of numbers to assess that
this transition stage had gone through smoothly.
If that's not the case, you can get some very odd results. Now, you can report
a maximum crack growth rate even if you've initiated cracking only over a tiny
portion of that transgranular fatigue pre-crack.
And this particular laboratory concerned, they actually were also using perhaps
the least suitable type of specimen, a very narrow DCB specimen of only ten
millimeter width.
So the bottom line is that if you only have a number saying, "I detected
two millimeters of stress corrosion cracking as maximum," you have no feel
whatsoever for how representative that is of the amount of crack growth rate
that actually took place during the test.
MEMBER LEITCH: Okay. Thank you.
MR. HICKLING: I mentioned I think earlier that all of these tests are obtained
in controlled primary water, and we paid a lot of attention to the fact that
we didn't have any off chemistry results in here and under two types of loading.
Just touching on one brief point which I'm going to eliminate, I hope, from
consideration straight way as well. We have an issue in that some laboratories
prefer to test using periodic slight unloading of the specimen, and what that
actually means here is nothing to do with simulating possible transience in
plant or anything at all. This is a typical way this is done.
It's a drop-in load to about 70 percent of the nominal value, usually about
once an hour during testing, and there are very specific reasons for doing that
which are connected with the way the test is conducted, and in particular, with
the way the crack growth rate monitoring equipment works.
It's an advantageous method of insuring accuracy of measuring your crack depth
on line during the test. However, there is a basic tendency if you start what
is ultimately some cyclic loading to accelerate crack growth because you'll
get out of a pure stress corrosion situation.
So we did some assessment of whether or not this would affect the results, and
the answer is that certainly for susceptible heats of material, it doesn't make
very much difference. It's possible that in less susceptible heats of material,
the application of this procedure may lead to slightly higher growth rates than
would otherwise have been measured. but we prefer to leave those in and accept
those because, again, it's a degree of conservatism.
Next one, please.
What have we got in this database with 158 points in terms of materials suppliers?
And this impacts directly on Peter Ford's discussion earlier about why we're
not considering material characteristics in the way that he would perhaps like,
and I think most of us would like to do.
First of all, we've got a number of domestic and overseas material suppliers,
and we've got 26 heats of material in the database with at least one screened
data point for heat.
The maximum number of heats we've got is 32 for any particular heat, and we'll
see a table a little bit later on which gives a little bit more information
on that.
What product forms? We've got a whole variety of product forms, thick wall tube,
forged bar, rolled bar, forged plate, and rolled plate. This is where the crunch
comes. Even for the materials which was used for the laboratory testing, the
information on the thermal processing history is extremely limited so that we
could not obtain the data we would have liked to characterize the material condition
in terms of its thermal processing history.
And of course, extrapolating to the field in terms of the nozzles that are out
there, that's an even worse situation. It's virtually impossible to get reliable
data on the thermal processing history of what is out there.
And the next slide, please? You're already there. Thank you
Which laboratories are involved? We ended up taking data from five laboratories,
one in the U.S. and four abroad who have done extensive testing on thick section
Alloy 600 material. They've done it at a whole variety of temperatures ranging
from 290 right up to 363 Centigrade, the desire, of course, often being to accelerate
the crack growth rate to reduce the testing time.
And since we know and have known for very many years that cracking PWSCC in
Alloy 600 is very highly temperature dependent, the first step was to try and
put all of this on a common temperature basis.
So we did that by choosing the most common test temperature, which is 325 centigrade,
or 617 Fahrenheit, and extrapolating everything back to that temperature using
an activation energy of 130 kilojoules per mole or 31 kilocalories per mole.
That is, more or less, the accepted activation energy for cracked growth rate
in this material, and even if you consider some of the more varied values that
have been obtained, the range for cracked growth data is actually pretty small.
It's from about 30 to 35.
So this does not have a huge effect on what we're doing.
MEMBER BONACA: I had a question regarding the previous slide actually. You said
that the thermal processing history of material is incomplete. I'm trying to
understand how significant. I mean this is Alloy 600. I mean, isn't Alloy 600
a pretty -- is it a common material we have or just specific to reactors?
MR. HICKLING: It's a common material in plants for milk processing and things
like that, yes.
MEMBER BONACA: Oh, okay. That's all I --
MEMBER ROSEN: Where it works rather well.
MR. HICKLING: It works extremely well. Alloy 600 was originally developed and
chosen because of its resistance to chloride induced transgranular stress corrosion
cracking. Its application in the nuclear field in the '60s and '70s originated
from that.
MEMBER BONACA: So what you're saying is that the thermal processing history
could be very different, I mean, depending on the application.
MR. HICKLING: Yes. Unfortunately we do know about the impact of the microstructure
on Alloy 600 cracking. We've known about that for many years. It's contrary
to what you would expect intuitively, particularly if you know about BWR stress
corrosion cracking because Alloy 600 works best when it has the most carbides
on the grain boundary, which is an initially surprising result in terms of --
so it's not chromium depletion phenomena.
So short of taking samples from every heat tested and actually doing a microstructural
analysis, it's very, very difficult to tie this one down.
Now, of course, in the lab you can do that. The problem arises if you have material
out in the field and you don't have archive material which is usually the case.
How do you ever get at what microstructure you're dealing with?
MEMBER BONACA: Thank you.
MR. HICKLING: How did we then go on to derive the curve? We knew, as I've just
discussed that the heat variation was likely to be very large in this data.
Our initial intention was to take a single heat of material where we had the
most data points and try and derive the dependence of crack growth rate on stress
intensity, on K from that heat alone.
Unfortunately by the time we'd rescreened all of this data, we simply did not
have enough data left to do that even for the heat where we had the most points
tested.
So we were forced to go back to an alternative approach, which is to adopt the
so-called Scott equation for this material, and the Scott equation was basically
developed quite some time ago using a very, very large amount of data on Alloy
600 obtained from steam generator tubing, which was undergoing primary water
stress corrosion cracking in the field.
So there's a huge number of heats, a lot of very susceptible heats, and a huge
number of data points in that original database, and that equation which was
developed originally in '91 basically says that the stress corrosion crack growth
rate is proportional to a constant alpha times the stress intensity nominal
threshold -- I'll come back to what we mean by that -- I'm sorry -- times the
actual stress intensity minus a value of nine, which is the nominal stress intensity
threshold to an exponent beta, which describes the basic dependence on stress
intensity.
And the Scott exponent from this analysis was 1.16.
The next --
MEMBER KRESS: Where does the erroneous relationship enter into the alpha?
MR. HICKLING: The erroneous relationship has been basically calculated in terms
of the alpha, yeah.
How does the data actually look in terms of what we're talking about here? These
are two examples from two different laboratories for two very different heats,
and in this particular case, at 325 degrees Centigrade, this is the Scott model
as defined by that equation developed from the steam generator tubing material.
And as you see, it comes down to very low crack growth rates, insignificant
crack growth rates at a nominal K of about nine, and this particular test is
producing data which clearly lies above that curve.
On the other hand, for some other material, a different heat tested in a different
laboratory at two different temperatures, this gives you some feel, incidentally,
for the temperature effect, there is the Scott curve for 290, and here is the
Scott curve for 325.
The data is falling below the curve at either temperature.
MEMBER WALLIS: That has nothing to do with the curve really, does it?
MR. HICKLING: Correct. You would be hard put to --
MEMBER WALLIS: It's a little low, but --
MR. HICKLING: One of the problems is that experimentally it's very difficult
to test over a wide range of Ks because you cannot get a big enough specimen
from the material available to test at high K values as you would like. So all
of the data tends to crowd between about 20 and about 40 megapascals.
MEMBER WALLIS: You can't really prove the nine because the crack worth rates
are so low down at that end.
MR. HICKLING: Absolutely, yeah. It's only a nominal threshold.
MEMBER WALLIS: -- a matching number then. If it's independent of temperature,
it's even lower. It's not magic.
MR. HICKLING: We actually considered -- at one point the expert panel debated
rather intensively whether or not we should try and make it zero or whether
we should make it four or six, and we did a sensitivity analysis. It doesn't
make a whole lot of difference because we're not using the result in that region.
We're not trying to describe initiation at all with this approach.
MEMBER ROSEN: That nine is not like Avogadro's number. It's not an important
thing.
MR. HICKLING: It certainly isn't.
(Laughter.)
MR. HICKLING: The true definition of a stress intensity threshold for stress
corrosion cracking is actually what you would get if you would decrease stress
intensity during a test and can prove unequivocally that the crack has stopped.
And in fact, that's a test which is almost impossible to do. So --
MEMBER BONACA: Why do you infer a curve like that, if I can go to the previous
curve?
MR. HICKLING: Yes.
MEMBER BONACA: I don't understand. You had a very specific curve that curves
and goes to 320 degrees, 330 to the right.
MR. HICKLING: Yes.
MEMBER BONACA: Or 325. How do you infer that curve from the distributional data?
You don't.
MR. HICKLING: Not at all. We can't. That is the point I'm making. We were forced
to go back to a curve which had been derived from a completely different database
and force fit it, if you liked to our data.
MEMBER BONACA: Right. I understand.
MR. HICKLING: Exactly right. So I've just covered that, but it's only an apparent
threshold, and we don't have data, but this is not going to be critical in use
because we're actually going to be at K values above, well above, say, 15.
There is another point that you have to mention. The threat exponent from the
steam generator tubing of 1.16 does imply a considerable dependence of crack
growth rate on stress intensity going right up, of course, to very high K values.
There's quite a lot of both field and test data which indicates this may not
be valid, that we may, in fact, be going too high at high stress intensities,
that there may be a plateau appearing.
But we couldn't convince ourselves that for our material that we had enough
data to draw a plateau.
MEMBER WALLIS: When it's high enough the material just breaks?
MR. HICKLING: Oh, yes. Eventually it would. You would eventually turn up where
you get the mechanical failure.
MEMBER WALLIS: How high is that?
MR. HICKLING: Much, much higher than anything we're dealing with, yes.
MEMBER APOSTOLAKIS: What are the typical K values you're going to have?
MR. HICKLING: You'll see when we come to the way this curve is being applied
we're talking typically about Ks in the range of 25, 30, something like that.
MEMBER APOSTOLAKIS: But if you subtract nine, that should have an effect, right?
MR. HICKLING: In what sense?
MEMBER APOSTOLAKIS: Well, the equation is DADT equals alpha K minus nine.
MR. HICKLING: Yes. The equation is just a fitting. The K minus nine is just
fitting.
MEMBER APOSTOLAKIS: Right.
MR. HICKLING: It was part of the original fitting to the steam generator data.
MEMBER APOSTOLAKIS: Are you going to use that equation again?
MR. HICKLING: Yes. That is the basis of--
MEMBER APOSTOLAKIS: So I don't understand why you say not critical for intended
use since the equation has a K minus nine factor there.
MR. HICKLING: I did mention that we, in fact, discussed extensively whether
it should be nine or six or four or even zero, and we tried out the effect of
plotting, replotting using all of those different curves, and in the region
of interest it makes virtually no difference at all.
It would make a lot of difference if you were trying to analyze the situation
at very low K values, but that's not where we are.
So we actually tried out the effect, and we stayed with the --
MS. KING: I would point out the third bullet here.
MEMBER APOSTOLAKIS: I guess it's because the exponent is just 1.16.
MR. HICKLING: Yes.
MEMBER SHACK: No, if you fit with zero, you'll get a different exponent. So
you'll change alpha and beta. So you'll get a different curve, but then if you
look at that curve between 25 and 35, they'll look sort of similar to --
MR. HICKLING: They'll more or less lie on top of it.
MEMBER SHACK: Yeah, the curves will move around a lot. You know, your alphas
and your betas will change.
MEMBER APOSTOLAKIS: Well, the alpha and beta change.
MEMBER SHACK: Yes, but the result in the range of 25 to 35 is not particularly
sensitive.
MR. HICKLING: Let me just repeat. Our original intention, our hope was actually
to fit our own data with the new curve, and that was the first approach adopted.
But unfortunately by the time we had screened out the reliable data points,
we just could not do it. We didn't have enough data over a wide enough range
of K. So the fall back position to this Scott curve is to some extent an artificial
one.
On the other hand, the Scott curve has stood the test of time, and it has been
used very widely, also for the analysis of --
MEMBER APOSTOLAKIS: Now, why didn't Scott have the same problem? Why didn't
he screen out inappropriate data?
MR. HICKLING: The main data base that Scott was working with were field inspections
on steam generator tubing of which there are literally thousands and thousands
of data points.
So he didn't have the same problems that we had. He had a huge number of heats,
far more than we have, and he had a huge number of tubes, which had been eddy
current tested. So he could determine differences in crack length and crack
growth rates. It's a quite different database he was dealing with.
MEMBER KRESS: The database you have, looking at, is crack growth rate versus
K.
MR. HICKLING: Yes, sir.
MEMBER KRESS: How did the various laboratories determine the K?
MR. HICKLING: That's a very good point, and I mentioned that the test methods
were different, constant displacement load. The simple answer, of course, would
be to use the standard equations, whatever form of pre-crack specimen they were
using in fracture mechanics, but the real issues that were involved are crack
front straightness, degree into which crack Ks change during the test, particularly,
for example on wedge open loaded specimens where the K value decreases.
And in one particular case, actually two French laboratories which produced
a lot of the data we're using, they went back without our prompting at the beginning
of this year and reevaluated their K values for every single specimen in terms
of remeasuring every specimen and recalculating.
MEMBER KRESS: Were these artificially made cracks at the start?
MR. HICKLING: Yes, the starter is always a fatigue pre-crack, which is a transgranular
pre-crack in the material. Now, that point in time you've got a pretty good
handle on what K is. It's later on as an irregular crack front develops you
have to consider that.
But that point was given a lot of attention.
MEMBER BONACA: Once you got the results at the end, did you ever go back and
took the 203 points that you threw away and see whether they would fit on that
curve?
MR. HICKLING: Well --
MEMBER BONACA: Would it be meaningful or just simply a meaningless exercise?
MR. HICKLING: I'm not sure whether it's particularly meaningful. I think you'll
see when we come to actually put up the curve in a second with the data, even
158 don't necessarily fit.
MEMBER WALLIS: We're all waiting for that with great anticipation.
MR. HICKLING: Yes, we'll get there very quickly.
The other point I mentioned is we have to take into account material heat variability
because we know how important it is, and we had a very limited number of options
as to how we're going to do that, and what we've, in fact, done is we've tried
to look at that in terms of calculating a different value of alpha for each
heat of material.
Now, what that means is we've taken every single heat of material, all 26 in
the database, and we've calculated the appropriate value of alpha to fit the
data for that heat to the Scott equation, and that would be the mathematical
formula.
No, go ahead. The formula is less interesting than this.
These are, in fact, the 26 heats of material from the different supplies. They're
rates in terms of the most susceptible in testing to the least from top to bottom.
You can notice, please, the differences in product form, which is implied here,
and notice also the difference in number of data points.
There is a certain tendency for laboratories to want to test a particularly
susceptible heat because it's an easier testing job, and in fact, that's why
some big numbers are coming up here, although there's quite a bit one down here
as well.
And there is also equally a tendency -- those heats where we have very little
data, and particularly ones where we only have a single data point are tending
more towards the bottom, the less susceptible heats where we have less cracking
observed.
And so you end up by doing this, by force fitting the Scott curve per heat,
you end up with a set of alpha values, the log mean power law constant, which
it varies, as you can see, between the most susceptible material actually we
had in the database, from six times ten to the minus 12, right down to two times
ten to the minus 13. It's quite a difference.
MEMBER WALLIS: Is it fair to ask what heat is? I don't understand what a heat
is in this context. Maybe I should have done my homework or something.
MR. HICKLING: What a heat of material is in this contexts?
MEMBER WALLIS: Yeah.
MR. HICKLING: It would be a single production lot as processed by the material
supplier.
MEMBER WALLIS: To do with heat?
MR. HICKLING: Yeah. It starts with the heat, yes. There are other factors involved.
MEMBER WALLIS: It's a production lot.
MR. HICKLING: Correct.
MEMBER WALLIS: It's not a property.
MR. HICKLING: No, not at all. It's just a material identifier. It's a number.
So we finally then get to where we want to go by taking the log normal fit,
the ordered median ranking of the alpha values for these 26 heats using standard
statistical methods.
I'm not myself a very good statistician. In fact, I'm a pretty awful one. Glenn
White, who did the data correlation exercise on this, and with a lot of input
from the gentleman on my right who has a very strong grasp of statistics, we
tried all sorts of methods, and I think this came out as probably the most valid
for looking at this database.
MEMBER WALLIS: So what you're saying here is that the properties of this stuff
are very dependent on how it was made.
MR. HICKLING: Correct.
MEMBER WALLIS: And that isn't a variable that's under control or is measured
in some quantitative way.
MR. HICKLING: Correct.
MEMBER WALLIS: So there's a tremendous amount of uncertainty about what's going
to happen.
MR. HICKLING: Yes. And that's why ultimately there's a limit to how far we can
go with a deterministic approach and why we have to get into a probablistic
approach.
But this is the result of doing this exercise. What we are actually plotting
here is the cumulative distribution of these alpha values for the 26 heats.
So every single point here represents one heat.
Now, it may have one specimen. It may have up to the maximum of 32 specimens
concealed in that calculated alpha value, and because it's a log normal distribution,
of course, it never completely goes to zero or to one. So as you can see, that
is this most susceptible heat which was identified, but our curve here is predicting
that you could have higher susceptibility heats and you could, in fact, have
very, very graphic cracking, which is ultimately going to be physically unreasonable.
There is a limit. It's very hard to define. There's no fully accepted mechanism
of Alloy 600 cracking. Therefore, it's very hard for first principles to calculate
a physically accepted maximum crack growth rate.
But we all know there has got to be one because otherwise you're getting electro-chemical
--
MEMBER WALLIS: What is your access there?
MR. HICKLING: This is the cumulative distribution of the alpha values as a function
of the actual values.
MEMBER WALLIS: What does that mean? You're just adding up the number of --
CO-CHAIRMAN SIEBER: It's the probability of this.
MR. HICKLING: Basically it's the probability function.
MEMBER WALLIS: But they all have different origins, and there are 27 tubes for
one alpha value, only one for another alpha value. I don't know how you get
a --
MEMBER APOSTOLAKIS: Are these points treated as being equivalent?
MR. HICKLING: Yes.
MEMBER WALLIS: But they're not.
MEMBER APOSTOLAKIS: Some of them come from a large number of test, some do not.
MR. HICKLING: Correct.
MEMBER APOSTOLAKIS: So shouldn't that be taken into account?
MR. HICKLING: Well, there's a limit to how you can do that. If you only have
one point to test, if the heat --
MEMBER WALLIS: You're looking for a pretty curve, and this looks quite pretty.
MR. HICKLING: No, no, it's not quite that. You're looking to try and represent
what you have. What you have is not what you'd like to have, but you're looking
to try and represent it in the fairest way possible.
And given the importance of material heat, we would have been much worse off
just taking all of the data and ignoring that effect.
Having said that, the full 158 data points for all of the heats feeds straight
into the probablistic analysis that Dr. Riccardella will be talking about. He
does not use this approach at all for that. He just takes the data as it comes
out.
MEMBER SHACK: Which has its own set of problems.
MR. HICKLING: Which has its own set of problems, too.
MEMBER APOSTOLAKIS: But still, you know, some of these points --
MR. HICKLING: Some have much bigger uncertainty than others.
MEMBER APOSTOLAKIS: Yeah.
MR. HICKLING: Correct.
MEMBER APOSTOLAKIS: And why is --
MEMBER SHACK: You can do the analysis estimating the uncertainties in each of
the alphas, and you find when you do that that the curve does not shift was
much as you would expect.
MR. HICKLING: We have gone through that exercise.
MEMBER APOSTOLAKIS: Now, why do we need one curve?
MR. HICKLING: Because we are trying to propose a single crack growth rate versus
K curve appropriate for dispositioning axial internal cracks in the field.
MEMBER APOSTOLAKIS: But why not a family of curves? I mean, I have uncertainty
here, don't I?
MR. HICKLING: Well, you don't have enough data to generate a family of curves.
Remember what we've done. We've --
MEMBER KRESS: Well, if you factor this probability in, you in essence have a
family of curves.
MR. HICKLING: Yes, you do in that sense, but you don't achieve very much because
your uncertainty -- I'm going to come on, if I may. Perhaps we could postpone
that question until I get to the applications slide as to how we intend to --
MEMBER APOSTOLAKIS: Why assume the data is constant and focus on the uncertainty
in alpha? I mean, do we really know, Peter?
MR. HICKLING: No, we don't know beta at all. Beta is assumed from this other
analysis. Beta has been adopted from an analysis from Scott.
MEMBER APOSTOLAKIS: But what alpha did scott use? He varied it?
MR. HICKLING: Yeah, the alpha value -- well, the definition of alpha depends
how you mean. On a heat to heat basis, yes. Alpha varies.
MEMBER APOSTOLAKIS: Beta doesn't change from heat to heat?
MR. HICKLING: No.
MEMBER APOSTOLAKIS: There is evidence that that doesn't happen?
MR. HICKLING: I'm not quite sure what question you're asking me here.
MEMBER APOSTOLAKIS: Why do you assume that Beta is constant?
MR. HICKLING: Because you can approach what you're trying -- you've got to remember
what you're trying to do. You're trying to define a crack growth rate which
is going to vary with stress intensity, first of all.
MEMBER APOSTOLAKIS: Yeah.
MR. HICKLING: There is no reason necessarily that we have to expect that the
material properties will affect the dependence on stress intensity per se. They'll
affect the propensity to cracking very much, but the actual stress intensity
dependence is no reason to assume that that should vary hugely between different
materials.
And, in fact, if you do the exercise that Bill is talking about, the fitting
to the individual heats and seeing how this curve moves, it doesn't move a whole
lot with the probabilities.
In an ideal world, you might only have one heat of material, and then you wouldn't
have this problem, but we're trying to tackle a very real problem here with
a larger number of heats out in the field.
MEMBER WALLIS: Well, it's a very strange way of doing things. If I understand,
you're looking at data from all of the different sources, and then you realize
there's a tremendous number of different alphas to correlate those data, and
then you are saying that we're not going to use some statistical thing to relate
to this to CRDM.
I want to know which one of these data points is most like our CRDM rather than
just taking a mean of a lot of things which might be something like it.
MR. HICKLING: Well, it's a good desire, but they all are. They're all from thick
section Alloy 600 material. They may just --
MEMBER WALLIS: There must be some reason that they're different by such large
factors.
MR. HICKLING: Yes, and the main reason is almost certainly the thermal processing
history of the material.
MEMBER SHACK: But if you had a CRDM nozzle picked at random, you don't know
whether it comes from the top of that curve --
MR. HICKLING: The middle or the bottom.
MEMBER SHACK: -- or from the middle or from the bottom, except on a probability
basis, that it's more likely to come --
MEMBER WALLIS: It's like testing a lot of nails from nail suppliers and measuring
something and then saying we're going to apply that to a bridge.
MR. HICKLING: But it's the standard situation you get into in stress corrosion
cracking where you're forced to use what's available, what you can generate
in terms of data, not what you would like to have, which is for every single
heat out in the field archive material with good quality data on it.
MEMBER SHACK: If you knew exactly what caused the spread, like the grain size
and the way they cooled it down, starting raw materials, you might be able to
go in and characterize a nozzle, but you know, that's asking a lot.
MR. HICKLING: There's a parallel here which is perhaps worth following very,
very briefly to a different problem in the BWR industry where stress corrosion
cracking has also been studied for very many years, also intragranular, but
where the mechanism of cracking has been tied down fairly well and has been
linked to exactly the sort of factors you're talking about so that you can tell
what difference potential makes, what difference material, what difference the
chemistry makes, and so on.
Unfortunately, despite 30 years or more of study, there is still at least three,
probably many more, credible mechanisms for primary water stress corrosion cracking
of Alloy 600, and so we do not have that in depth understanding at a fundamental
level to do that.
MEMBER KRESS: Yeah, nd I think the only recourse is to fall back on a probability.
MR. HICKLING: So where does this get us to? Let's come back to that Christine
and just throw up what this actually does.
These are the 158 data points. As I remind you, each one is plotting growth
rate in the test against the representative K value for the test, and again,
you will notice the bunching between the 20 and 40 values of K, just the odd
ones which are higher or lower.
This is the modified -- this Scott curve, called the modified curve, but that's
--
MEMBER APOSTOLAKIS: This curve has nothing to do with the previous curve?
MR. HICKLING: Yeah.
MEMBER APOSTOLAKIS: Yeah, what?
MR. HICKLING: This curve is calculated.
MEMBER APOSTOLAKIS: Okay.
MEMBER WALLIS: But the naive observer would say that the curve has nothing to
do with the data whatsoever.
(Laughter.)
MR. HICKLING: Possibly true, possibly true.
CO-CHAIRMAN FORD: But the MRP curve, John, is the mean curve from the previous
graph. It's using the alpha mean.
MR. HICKLING: What I'm going back to, it does, of course, have -- if we could
just go back to the previous slide.
To get to that curve, we -- let's go back to the curve with the alphas, please.
Thank you.
You're basically given the choice here. Once you've determined this dependency,
how do you handle the uncertainty, and what value of alpha are you going to
use to plot your single curve? Because you need to end up with a single curve
in order to do anything sensible in the field.
The value that we've chosen is to use the 75th percentile from this curve for
our value of alpha, and this is, in fact, the mean, if you like, of the upper
half of the distribution. So it's not the median value here. It's considerably
higher than that. There's a reason for this. It's basically that we are trying
to make a best estimate of lightly cracked growth rate in the field, and there's
obviously no point in going unrealistically low, but there's no point either
in going absolutely unrealistically higher for every single heat of material
that's out in the field.
The conservatism that you might want to apply, we feel should be added later
in the process when you're evaluating and dispositioning an actual crack, and
you have plenty of opportunity there to add engineering conservatism rather
than adding it in a hidden form at this stage in the data.
And the ASME code gives some basis for this approach of taking the 75th percentile.
So this is how we define the value of alpha here that we use when we create
that next curve. Okay?
MEMBER APOSTOLAKIS: So this curve then is the Scott curve with alpha equal to
this value, the 75th --
MR. HICKLING: No --
MEMBER APOSTOLAKIS: -- beta equal to 116?
MR. HICKLING: The shape is modeled entirely on the Scott curve. So the exponent
is derived from the Scott curve, and the nominal threshold is derived from the
Scott curve.
MEMBER APOSTOLAKIS: And alpha, too.
MR. HICKLING: No, the alpha is derived from our actual data.
MEMBER APOSTOLAKIS: Yeah, but in this plot it's the 75th percentile of the previous
curve.
MR. HICKLING: yes.
MEMBER APOSTOLAKIS: Okay.
MR. HICKLING: But that previous curve is for our own data on the thick section,
not for the steam generator.
MEMBER SHACK: But isn't the MRP curve the 75th percentile? The modified Scott
was an earlier curve that had been proposed.
MR. HICKLING: Yes, yeah. The MRP curve is what we calculate on that basis.
MEMBER KRESS: Now, the data points --
MR. HICKLING: And it lies -- it's parallel to obviously the Scott curve because
it takes the shape from it. It's force fit to it, but it's about 20 percent
higher.
MEMBER KRESS: Yeah, but the data points on this curve are the same data points
you use to get your probablistic alpha. So it's no surprise that it kind of
goes through the mean of them because the 75th on that cumulative is like a
mean.
MR. HICKLING: Yeah, it's the mean of the upper half.
MEMBER KRESS: So it's just reflecting the previous curve when you see it do
that.
MEMBER BONACA: And I hope the Scott curve had a better fit to data than this.
MR. HICKLING: Well, that's why we used it.
MEMBER KRESS: Well, all this is saying if you go back to that previous curve,
it went from ten to the minus 13 up to ten to the minus 11, and you look at
the data on this curve. It does the same thing. It's a reflection of this curve
right here.
MR. HICKLING: That's right.
MEMBER WALLIS: And any theory that you had that you forced alphas to be like
this would go through the data.
MEMBER KRESS: Oh, yeah, absolutely, because you forced it to go through the
data. And you forced it to kind of go through that part of the data.
MEMBER WALLIS: Yeah. That conclusion is Scott is wrong. I mean, Scott has nothing
to do. Scott --
CO-CHAIRMAN FORD: Scott can't be wrong because it's based on --
(Laughter.)
CO-CHAIRMAN FORD: I'm not saying a Scott can't be wrong. But the Scott curve
is an empirical relationship based on field data.
MR. HICKLING: Yes.
CO-CHAIRMAN FORD: But what I'd like to know John is you choose the 75 percentile
of alpha according to the MRP curve.
MR. HICKLING: Yes.
CO-CHAIRMAN FORD: But I know there was some data where you should be at the
95th percentile. What was the reasoning behind the choice of 75 over the more
conservative 95 percent?
MR. HICKLING: The reasoning is, Peter, quite simple, that we feel that in screening
the database we've already applied quite a considerable amount of conservatism.
There are a lot of material, as you know. For example, we couldn't consider
any heats which didn't show cracking at all. So they're eliminated.
The reasoning is quite simply that we feel that this curve is a good representation,
if you like, a conservative representation already of what is actually out in
the field.
There will be a lot of heats out in the field which will crack at very much
lower rates than this, and I'm going to come onto a comparison with field data
in the next slide.
MEMBER KRESS: You're saying that all the data you threw out would fall below
that curve on this plot basically.
MR. HICKLING: In general, in general. There are two types. That would be a little
bit too general, that statement. We threw some data out, for example, because
it was tested in off chemistry, and that might have been higher, but a lot of
the data we threw out would have quite clearly fallen well below this curve.
For example, in some of the wedge overloaded data which we threw out, those
points were coming out at least an order of magnitude lower than they probably
should have been simply because of problems of artifacts of testing.
MEMBER APOSTOLAKIS: But if you use the 75th percentile of alpha, wouldn't you
expect most of the points to be below the curve? That doesn't seem to be--
MR. HICKLING: No.
MEMBER APOSTOLAKIS: -- the case.
MR. HICKLING: It depends entirely on the distribution.
MEMBER KRESS: That distribution, the 75, is actually close to the mean really.
MEMBER SHACK: Well, it's the 75th percentile on the heat. Now, if a susceptible
heat has 32 data points, it's going to skew. When you look at data point by
data point, it skews the distribution, which is one argument for doing it by
heat. Otherwise you overly weight --
MR. HICKLING: Right.
MEMBER KRESS: And so this is a log scale down here
MEMBER APOSTOLAKIS: Wait, wait, wait. I'm speaking of the 75th percentile of
this curve, right? If I plotted these points, you know, in the next curve, then
I should have most of them below the curve.
MEMBER WALLIS: Yes, but you didn't.
MEMBER SHACK: But you didn't.
MEMBER APOSTOLAKIS: But you didn't.
MEMBER SHACK: You plotted the raw data.
MEMBER APOSTOLAKIS: You plotted the raw data, which now brings you back to the
earlier assumption of using these points as being equivalent. Doesn't that tell
you something about the uncertainty of each point and how important it is?
The fact that the new curve doesn't seem to be on the high side probably tells
you something about the --
MEMBER WALLIS: No, it tells you there were 21 points for heat one and only one
for heat 26.
MR. HICKLING: But there's a strong tendency for the laboratory to have tested
a susceptible heat if possible. That's true in the whole history. They don't
want to get a zero result which is of no use to anybody.
So there is an innate bias in any stress corrosion cracking test data to have
chosen usually the most susceptible material they could get their hands on at
least initially.
MEMBER SHACK: But the question is: do you want to characterize the variation
in the set of test data that you have or in the population of heats of material
that you're likely to encounter in the field?
If you want to characterize the variation in your test data, you do your statistics
on all of the data points. If you want to do that, except you sort of hope that
you have enough data that's really characteristic of the population.
MEMBER APOSTOLAKIS: Let's go to the next curve.
MEMBER SHACK: But what you're looking for is the population.
MEMBER KRESS: Well, why did you feel like you had to not use the whole curve?
If you do a probablistic fracture mechanics, you could have used that whole
distribution.
MR. HICKLING: We are doing it. The probablistic fracture mechanics uses the
whole database and --
MEMBER KRESS: Okay. I feel better about it then.
MEMBER APOSTOLAKIS: So we could have a family of curves here, you know, with
some confidence instead of a single curve, and that's what you're going to do
in the probablistic --
MR. HICKLING: Exactly, except the probablistic, as I say, is not based on the
MRP curve at al. The MRP curve we're trying to achieve is a reasonable representation
of what we would expect for crack growth rate already involving some conservatism
for heats out in the field.
MEMBER APOSTOLAKIS: So this is a reasonable representation?
MR. HICKLING: As Bill says, of the heats that are likely to be out in the field.
MEMBER SHACK: It's his choice.
MR. HICKLING: This was the expert panel's recommendation.
(Laughter.)
MEMBER SHACK:
MEMBER SHACK: Yet in a deterministic world you pick one curve. Which curve do
you want to pick?
They have chosen the 75th percentile for the reasons that John has stated. You
could make arguments that it should be the 95th percentile. You want to bound
all of the data. You could make it the 50th percentile. You want a representative.
You know, you have to decide in a deterministic world with a lot of scatter.
You have to make an argument for which curve you want to pick.
MEMBER APOSTOLAKIS: And the argument is that the points above the curve don't
matter that much?
MR. HICKLING: Well, let's develop the argument a little bit more because the
test of any curve is does it describe the field observations, and that's the
point. It's already indicated a little bit.
There are actually two points written in here, which I'm going to come onto
in the next slide what those are. There is very little data available in the
U.S. from the field on nozzle cracking where there have been sequential measurements
of crack length and depth.
The only data that's available is from one nozzle in D.C. Cook 2 where a crack
nozzle was allowed to operate for a certain period of time, and there was increase
in the measured length and depth of the crack.
And these two points are plotted here. This is the length increase of that crack,
and this is the depth increase.
Now, agreed this is only one isolated indicate, but it is worth noting that
both of those points fall very well below that curve.
We go on to the next slide --
CO-CHAIRMAN FORD: Could I just interrupt for one minute, John? I wanted -- this
is the reason why we are discussing this data. This is one of the first times
that this group has seen these data, and I wanted to be aware of the amount
of work that's gone into this area.
However, we could go on forever discussing this, and what I would like to suggest
is that we will finish this at 12 o'clock, this particular presentation at 12
o'clock. We will recess for lunch for three quarters of an hour, and we'll come
back at quarter to one, and that will give Glenn hopefully time to do his presentation
and leave when he wants to do. Yes?
MR. WHITE: Yes.
CO-CHAIRMAN FORD: Will that be okay?
So, John, could you pick and choose and try to finish by --
MR. HICKLING: Yes, we can get through the rest very quickly, I think.
(Laughter.)
CO-CHAIRMAN FORD: Yeah?
MR. HICKLING: With your help. Basically whenever you do a comparison from what
you derive from the laboratory data with the field data, we've talked a lot
about the uncertainties in the laboratory data, but it's worth remembering that
there are very considerable uncertainties in the field data because we're basically
talking about differences between two ultrasonic measurements of crack size,
and we are really analyzing the difference between the delta between those two
measurements.
So there's considerable NDC uncertainty, feeds in straight away here.
Secondly, there are uncertainties in the estimates of K, depending on how you
analyze the residual stresses for the particular component concerned, and that's
a very significant problem in this area.
And, thirdly, of course, there may be some uncertainty in the actual operating
temperature of the nozzle, and we know how corrected these values are for temperature
in different plants and in different countries.
I've showed on the previous slide the D.C. Cook data. The main body of field
data we have available to compare with our curve is, in fact, French data because
the French, once they detected cracking Bouget, did a lot of ultrasonic inspection,
and they never had a second leakage.
So there is a lot of field data out there, and we made very considerable efforts
to obtain everything we could.
The French reported their data at certain operating temperatures for their plants,
and there has been some movement in what they've reported over the years as
the operating temperature of different plants.
We have taken the latest report we were able to obtain on individual plants
and extrapolated the reported data to a common temperature of 325 degrees Centigrade
in order to compare it with our curve.
What we've done, rather than just comparing it simply with the curve, is we
decided to go to a statistical approach here to show you how, in fact, the data,
the screened data in our database, is going to work. And what we've actually
done for the comparison is the following.
For every point where we had a field data point at a particular K value where
we could derive a crack growth rate. We've done some random sampling from the
upper half of the MRP distribution of crack growth rates, are using the same
approach that we got, basically the letters to the 75th percentile, and using
the K dependence of the Scott equation.
Let's just put up the results, and then we can come back to that. In this diagram,
the black points represent the EDF field data extrapolated to the nominal temperature,
325, from the reported temperature of the head.
The red data points are the data points obtained from our MRP distribution applying
this Monte Carlo approach to the top half of the distribution. So every time
you did that, you'd get a different set of red points.
But remembering the uncertainties in that field data, we feel that's a more
valid comparison than just putting a curve through it, and at first glance you
can see that the Monte Carlo does produce some very high crack growth rates,
of course, as you'd expect from the MRP distribution, and the agreement doesn't
look that bad.
In fact, the next curve shows what that would look like on a cumulative probability
plot of the French field data here, the black points, and this statistical treatment
of the upper half of the database, which are the red points.
And there's no denying the French field data is higher, showing that the cracks
measured in France in the field did grow more rapidly than what we're predicting,
and when we consider there are very real reasons for that, as Larry mentioned
earlier, we don't think it's just a matter of chance that the French plants
have this problem so much earlier.
MEMBER WALLIS: Would you do the exercise of taking random numbers between 20
and 50 for K and between 1E minus 11 and 1E minus nine for crack growth rate?
Just take random numbers, do exactly the same thing you've done here. You'll
get the same sort of picture.
MEMBER KRESS: Well, that's what he did, except the random numbers are --
MEMBER WALLIS: But what does it tell me? If the random numbers give the same
result as your data, I'm not quite sure I've learned anything from the data.
MR. HICKLING: No, they're not entirely random numbers. It's a Monte Carlo treatment
of part of the data.
MEMBER WALLIS: Well, no, I mean if I look at this curve here with this distribution
of points.
MEMBER APOSTOLAKIS: Which distribution are we referring to? I haven't seen a
single distribution here.
MEMBER WALLIS: If I had random numbers here, I get the same --
MEMBER APOSTOLAKIS: Which distribution? Of the alpha?
MEMBER WALLIS: The alpha.
MEMBER KRESS: Yeah, but they only selected from the top half though.
MR. HICKLING: Correct. It's an attempt to recommend the sort of variation that
is inherent in the data, whether it be from the lab or the field.
MEMBER WALLIS: But there are people who have tried to publish reports like this,
which show that taking random data on the same graph gives the same result,
and that doesn't give me a good feeling at all that it's a useful exercise.
MEMBER KRESS: Well, it's a way to compare the French data to this database that
went into making the curve. That's all he's saying. It's a way to compare those
two.
MEMBER WALLIS: But if you compare the random numbers thrown at the --
MEMBER KRESS: But he's showing what would happen if you took the French data
and put it on this same curve with -- you'd have ended up with a different distribution.
MEMBER APOSTOLAKIS: The French data were not part of the derivation of the curve
for alpha?
MEMBER KRESS: No, and they say it's clearly a different set of data, and they
have reason to believe it should not be part of the database, and I think that
reason is maybe weird, and that is, well, they started cracking a lot earlier
than ours. So it must have been something.
MR. HICKLING: No, no. Excuse me. There are two separate issues here. That is
the reasoning why the French data will always come out higher no matter how
you treat it, because we do believe that the material susceptibility was higher.
The one thing we do know is that the material processing temperatures in general
were much lower in France for that nozzle material, and there's good reason
to expect that that would lead to a higher degree of susceptibility.
The second point, the reason why we didn't use the French data, for example,
in deriving our curve is that there are uncertainties in the French field data
which we cannot fully tie down and which we are ultimately somewhat unhappy
about. We've extrapolated up very much in temperature. Whether or not that's
fully justified is another issue, and it's an issue we couldn't solve.
MEMBER KRESS: It depends on whether your final product you want to be highly
conservative or you want to be a representative value, I guess.
MR. HICKLING: Exactly, and the feeling is that we are trying for a representative
curve, and the conservatism which needs to be added is added in the engineering
analysis later on and is visible, not hidden in some way.
MEMBER KRESS: Yeah, which the ACRS has said is the way you ought to do things
with respect to different issues in the past.
We have always advocated that as the right approach.
MEMBER APOSTOLAKIS: Well, yeah. There was a bullet that said if you did something
because it was conservative. I mean, they're not as pure as it would seem.
(Laughter.)
MEMBER APOSTOLAKIS: Right?
MEMBER KRESS: There's always a mixture.
MEMBER APOSTOLAKIS: Yeah.
CO-CHAIRMAN FORD: It was the screening criteria which they said was conservative,
and that's why they're using the 75th rather than the 95th percentile for alpha.
It's reasonable.
MEMBER APOSTOLAKIS: So what Tom said is not quite accurate.
MEMBER LEITCH: John, one thing that concerns me regarding that French data,
I guess, I've always wondered whether -- you know, we spend a lot of time talking
about crack growth rate. I'm wondering about the depth of the crack at initiation.
In other words, does the crack grow more or less linear? It's how many inches
per year from zero, or might it be a fact that instantaneously the crack proceeds
to some depth?
MR. HICKLING: No, definitely not instantaneously. You're quite correct. We're
not trying to describe that whole phase of initiation and early growth, but
all that we know about both primary water stress corrosion cracking in general
suggests that the initial phase of crack growth is very, very slow, indeed,
and getting the crack -- remember in the field we're not dealing with transgranular
fatigue pre-crack which then goes into granular at all. We're dealing with a
crack which develops as an intragranular stress corrosion crack at a point in
time where you can't calculate it.
And all of the evidence is that a huge part of the lifetime, perhaps as much
as 85 percent of the lifetime of the crack, as it were, is developing the initial
crack, whatever you'd like to call initiation, and growing it to a level at
which you can detect it with NDE methods.
So we're not addressing that whole area at all here. We're just saying what
would we do to disposition once we find a flaw which is large enough to be found
by NDE.
And I think the one thing that you can be sure about is that there's nothing
instantaneous about stress corrosion cracking in that sense.
MEMBER LEITCH: So you're saying that the evidence seems to suggest that that
initial phase is relatively slow compared with the ongoing. I was wondering
if -- you know, in my mind I had pictured a model that was just the opposite
of that. Initially it took a quick depth and then the growth was slow from there.
MR. HICKLING: No. I think you'd find pretty uniform agreement among anyone who's
worked on stress corrosion cracking.
MEMBER APOSTOLAKIS: So the growth rate is independent of the size of the crack?
MR. HICKLING: No, it's actually not. It's very dependent upon it.
MEMBER KRESS: It's part of the K
MEMBER APOSTOLAKIS: Oh, K, K.
MR. HICKLING: It's later part of the K, and in the very initial stages, it's
more complicated that --
MEMBER APOSTOLAKIS: Now, where do the curves cross up there? Is there any reason
why they should do that?
MR. HICKLING: Yes, because the black points, in fact -- well, it's a function,
of course, of the sampling that has been applied to the MRP distribution to
get these particular set of points, but if we just go back very quickly to that
alpha curve, it's a point I'd like to make.
Remember this is a log normal fit which is approaching one exponentially. So
you are predicting infinitely high crack growth rates, albeit with a very, very
low probability that it will ever occur. So that is physically unreasonable.
And, in fact, as Dr. Riccardella will talk about in the probablistic talk this
afternoon, for that purpose you're going to have to truncate this log normal
distribution, go to a log triangular because it's physically unreasonable to
go to infinity. A stress corrosion crack would never do that. It can't do.
But the effect of using it in the way we've just done it is, of course, it can
generate some very high crack growth rates even at low K.
MEMBER APOSTOLAKIS: So how big was your Monte Carlo sample? Was it big enough
to pick up those values, the sample?
MR. HICKLING: Yes.
MEMBER APOSTOLAKIS: You said you did a Monte Carlo.
MR. HICKLING: You mean the number of iterations?
MEMBER APOSTOLAKIS: Yeah, yeah, because it will be a very large number to start
picking up the very unrealistic --
MR. HICKLING: No, I'm not saying we'd be picking up any which are way out in
the table here, but I'm saying it's inherent in the approach that we're using.
MEMBER APOSTOLAKIS: Does that explain why the curves cross?
MR. HICKLING: I think so, yes, because the French field data is real data, albeit
with uncertainties.
MEMBER APOSTOLAKIS: Oh, okay. So it's an artifact.
MR. HICKLING: Can we go on quickly?
I want to make one very -- one before, please. Thanks -- I want to make one
very important point. In actual fact, in France in the regulatory context, the
French finally did not use any of these approaches. The actual French approach
that was finally agreed upon was that in no case did the actual measured crack
growth rate in the through wall direction of any crack which was found in plant
exceed four millimeters per year, and this was actually the figure they adopted
irrespective of head temperature as a limit which would allow them to justify
continued operation for at least one cycle even with cracks which were already
11 millimeters deep.
MEMBER WALLIS: And your French data plot shows 20 or 30 millimeters a year.
That's in the other direction.
MR. HICKLING: No, but that's because it's been temperature corrected, and it's
been pushed up a lot in temperature. The reported temperatures for the French
plant, as I said, have moved somewhat, but they tended to move down quite low.
So we've had to extrapolate up an awful lot, and we're not very happy about
having done that, quite frankly.
MEMBER WALLIS: That's one reason they're so high.
MR. HICKLING: Absolutely.
MEMBER WALLIS: Or it is the reason they're so high.
MR. HICKLING: So moving on to what do we actually intend to do with this curve
and why do we think it makes sense, it's intended that it would be used to detect
the disposition, PWSCC floors, either if they're axial ID floors or if they're
below the J-groove weld, i.e., we're not -- floors which are not part of the
pressure boundary.
The main application we see is a deterministic evaluation of axial ID floors
which are part of the pressure boundary. We're not intended to use it, as we
discussed earlier, at very low K values. Such floors, once detected, will already
be well above any K value that you might be looking at here.
And this is to give you a feel for a generic calculation of what that ID axial
crack growth would look like. The Y axis here is showing the depth of the axial
ID crack initially, and this is showing the calculated operating time to reach
a 12 millimeter deep crack, which would be 75 percent through wall acceptance
limit in the nozzle, to give you a feel for the sort of way in which this would
pan out.
MEMBER WALLIS: There's a lot of uncertainty in this, isn't there?
MR. HICKLING: There's a lot of assumptions I would say is perhaps a better word
as well rather than --
MEMBER WALLIS: I don't know how you can get one curve from that tremendously
uncertain data without showing many curves or something.
MR. HICKLING: Well, the way we get to a single curve is defined because of the
way we've defined the curve. I think the question is what uncertainty remains
in the analysis.
For example, we've assumed in this particular case a particular K value based
on a residual stress here. Now, this is a generic calculation. It's purely an
example calculation, nothing else.
In any application of this, we'd expect that a found floor would be dispositioned
correctly in terms of the best possible stress analysis to reduce the uncertainty,
for example, in --
MEMBER KRESS: Yeah, but I would have also expected for a specific case for the
decision maker to make an appropriate decision, you would have a set of curves
for the distribution of the uncertainty about that curve.
MR. HICKLING: Not in terms of the deterministic approach, no.
MEMBER KRESS: Well, yeah, but that's one of our problems with the deterministic
approach. We never know what the uncertainties are, and the uncertainties are
what drive our decision making process.
You know, if that curve had uncertainty bounds on it, five and 95 percentile
or something, then as a decision maker I'd have enough information to at least
think about what decision I want to make, and you could do that with the database
you have. It's inherent in it.
Pardon?
MEMBER APOSTOLAKIS: We're going to discuss this this afternoon.
MEMBER LEITCH: I'm afraid I don't understand how that curve would be used. Maybe
I don't understand the axes.
MR. HICKLING: In an actual plant situation, you would detect with NDE a crack
which you would size as best as you possibly could.
MEMBER LEITCH: Right.
MR. HICKLING: And here we're saying that we might size it as let's take an example
and size it at four millimeter depth (phonetic).
MEMBER LEITCH: Okay.
MR. HICKLING: You do the best possible analysis you could of the residual stress
driving that crack based on all sorts of things, including nozzle downhill angle
and all of the other things you might be able to put into that to get your K
value, which would feed into the equation here.
You'd adjust your head temperature to the correct value for the actual plant,
and you'd then read across and determine that without adding any subsequent
conservatism, which you would almost certainly want to do; the prediction from
the MRP crack growth rate curve would be perhaps in that case that you would
need something like 16 months or 15 months for that crack to have grown from
four millimeters deep to 12 millimeters deep.
MEMBER LEITCH: Okay.
MEMBER KRESS: And that's part of the analysis you make to determine whether
you can continue operating in a certain amount of time.
MEMBER BONACA: So that curve will shift.
MEMBER ROSEN: Well, if you have an 18-month cycle, you look on that curve and
see if your operating time is greater than 18 months and it says it is; then
you can run the cycle.
MEMBER BONACA: Right.
MR. HICKLING: That's a good point. This is the temperature of what we regard
as the hottest head, which might be actually applicable.
MEMBER APOSTOLAKIS: But is there any reason to believe that this curve is conservative?
I mean, in a deterministic world at least you want to have something conservative.
Is it conservative?
MR. HICKLING: There are some conservatisms inherent in the derivation of the
curve. That's the point I was trying to make earlier. Whether or not it's a
conservative curve is a global question which is very difficult to answer.
We consider that it's a representative curve for some of the heats which are
more likely to crack because remember it's the 75th percentile, not the 50th,
of our database.
Could I just go quickly over the very final slide?
There is no intention, I think, in the industry to try and disposition OD cracks
which are actually found. Going back to what we talked about right at the very
beginning, if we were talking about hypothetical calculations, we would recommend
that this factor of two, which represents the uncertainty in the chemical environment
be put onto that curve.
And a subgroup of the experts did look at the experience. We still think the
arguments as I mentioned that we put forward on the environment are valid in
the non-Davis-Besse situation, which we consider to be the usual case which
has been found to date.
However, last slide Christine.
It wouldn't be valid, and we're not claiming that it would be if the leak rates
were sufficiently high to get a large, local decrease in temperature, cavity
formation, and steel.
That brings up the question: what would happen with stress corrosion cracking
of Alloy 600 in that case?
And that takes me back to this point I mentioned earlier, that in general, we
think of Alloy 600 as being very resistant to cracking in acid media. There's
very little data available. What there is shows that in order to get cracking
in concentrated boric acid, you need quite high levels of both oxygen and chloride
contamination, not just one or the other.
And interestingly, the effects at N was at intermediate temperatures, suggesting
that we're now in a different type of Alloy 600 cracking, not the primary water
stress corrosion cracking we've been talking about.
And that's all I had.
CO-CHAIRMAN FORD: Thank you very much, John.
MEMBER KRESS: The factor of two that's put on there, because of chemistry uncertainties,
strikes me as being a little strange in view of the uncertainties in the data
about getting the curve in the first place. It's just overwhelmed by the --
MR. HICKLING: It's handling a different situation.
MEMBER SHACK: It moved the whole population is the theory.
MR. HICKLING: Yes.
MEMBER SHACK: On any crack growth rate of heat, it's insignificant compared
to the variation between heats, but if you're moving the whole population.
MEMBER KRESS: I'll have to think about that one. I still think it's gilding
the lily.
CO-CHAIRMAN FORD: John, thank you very much indeed.
I'd like us to go into recess until quarter to one when we'll start again. Quarter
to one, guys.
(Whereupon, at 12:04 p.m., the meeting was recessed for lunch, to reconvene
at 12:45 p.m., the same day.)
A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N
(12:49 p.m.)
CO-CHAIRMAN FORD: Okay. We're back in session.
We're going to start with the technical assessment of Davis-Besse's degradation.
Am I correct?
MS. KING: Yes, you are correct. I do have both presentations for you, and in
your packets, this would be Slide 81, about three quarters of the way back.
And we will come back to the fracture mechanics.
MS. WESTON: If I may, some of the slides and tables are in your book starting
at page 131.
MEMBER APOSTOLAKIS: So when you said the 81?
MEMBER KRESS: The package of slides.
MEMBER APOSTOLAKIS: This package, yes. Okay.
MR. WHITE: Good afternoon, everyone. My name is Glenn White, and I'm with Dominion
Engineering.
Since March 22nd, Dominion Engineering has been supporting the Electric Power
Research Institute and the Materials Reliability Program on assessing the Davis-Besse
experience. Specifically we've been trying to understand, based on calculations,
analysis work, and also looking at experimental data that's available, what
the degradation progression was at Davis-Besse.
MS. KING: We're in animate mode. Let me fix it real quick. Go ahead.
MEMBER APOSTOLAKIS: Show without a dimension. See that on the left at the bottom?
MS. KING: Thank you very much.
There we go.
MR. WHITE: Okay. The presentation that I have prepared that's in the packet
here is approximately 15 slides of material that summarizes the various mechanisms
that could possibly be active and summarizes our conclusions as to what we believe
happened at Davis-Besse, what the likely progression of degradation was.
Two weeks ago at an NRC meeting with some of the NRR staff and research staff,
I presented a longer presentation, 63 slides. That presentation is available
on the NRC Web site, and we have that as back-up material for this discussion.
So if there are questions that get into particular areas, I'm prepared to answer
them using that longer presentation, but the original time allotment for my
talk was only a half hour. So that's why they're sticking to the 15 slides in
the packet.
MS. WESTON: Glenn, they have copies of that package in the notebook.
MR. WHITE: Okay.
MS. WESTON: They have the whole package.
MR. WHITE: Great, perfect.
I'm going to start off talking about the purpose of this work, the approach
that is called for, and then get into the individual mechanisms briefly, as
I said, and then outline what the likely degradation progression was based on
our analysis work, supplemented with experience and experimental results, and
then also touch on the most relevant experimental test that had been performed
in the past because I think it's important to touch on that.
We've done work to try to quantify the chemical environment and the thermal
hydraulic environment along the leak path in the annulus on the OD of the nozzle,
and so there are a lot of other analyses that we can get into, as I say.
So if we go to the next slide, the purpose here is to answer two main questions
that have been put forth. The first one is if there is significant degradation
it will be detectable visually, by doing a visual inspection of the region above
the head.
And the unit could be detectable a couple of different ways. One, you might
see a void directly so that you could see the wastage directly.
But the other way, you could infer that there might be wastage that would require
a closer look if you found a significant amount of deposits, either boron deposits
or some corrosion product deposits. So that's the first main question.
The second main question has been put forth is what is the time scale of this
process following initiation of a through wall leak. Is there a period of time
that we all have assurance that we can't reach unacceptable wastage? That's
the second question.
A related question to that is: what is an unacceptable level of wastage, and
I'm not directly addressing that in this presentation here because it's a closely
related, but a slightly different subject that really goes to the structural
stress calculations.
What I'm going to be concentrating on is the degradation progression, the environment
in the annulus, and the various corrosion and potentially erosion mechanisms.
But on the question of what is acceptable, I will mention that in the early
'90s, in the '93 time span, the three owners' groups did finite element analyses
taking out a certain volume of the -- actually six cubic inches of volume of
the low alloy steel head, and they did that using different geometries of the
assumed loss, different aspect ratios of the voids.
And at that time it was determined that six cubic inches allows the code margins
to be maintained.
MEMBER WALLIS: It depends how it's removed.
MR. WHITE: It depends how it's removed, but each owners' group took two or three
different bounding assumptions. So based on those --
MEMBER WALLIS: But if it's a straight hull, it's very different from taking
off six cubic inches all the way around.
MR. WHITE: Yes. For example, it would take all six cubic inches along the other
surface, the top surface of the head, or you could take the six cubic inches
along the bore, and no matter how they were taken out, the stressors are still
within code margins.
MEMBER WALLIS: That's assuming you had a lined head?
MR. WHITE: Assuming different geometries, different bounding geometries is what
they did.
Since that time, we have just recently begun to look at this question of what
is acceptable wastage, and Dominion Engineering has performed some preliminary
finite element analyses, taking out some of the elements that make up the head,
and the conclusion from that work is that it's most likely significantly more
than six cubic inches can be lost and still the primary membrane stresses will
still be below the code allowable stress intensity values.
And just mentioning because this is a related question --
MEMBER WALLIS: This is with the stainless steel liner, cladding?
MR. WHITE: The cladding is a second question. The first thing we did --
MEMBER WALLIS: Well, without your cladding you could make a hole six cubic inches,
couldn't you? You could drill a hole through it and remove six cubic inches.
You have a small LOCA that's all
MR. WHITE: We've also looked at the issue of the cladding, and I believe later
this afternoon there will be some discussion about the margins in terms of the
cladding for Davis-Besse, and there, again, there could be a significantly large
area where the cladding is retaining the pressure.
MEMBER WALLIS: Okay. So like 200 inches at Davis-Besse?
MR. WHITE: Yes, approaching 200 cubic inches of material loss at Davis-Besse,
and I'm going to put that in the context of the progression in some other slides
here.
Okay. The basic approach is to examine how the various conceivable mechanisms
and material loss change as the leak rate increases. Through our analysis work,
what we found is it's really the rate is the controlling parameter for two main
reasons which are shown down here.
Number one, the level of cooling. When you start with primary water, it has
a certain enthalpy, about 613 BTUs per pound. If you have saturated steam at
atmospheric pressure, its enthalpy is higher. So you need to have some heat
input in order to completely boil off that primary water.
But the primary water because of the temperature and the pressure, it does have
enough enthalpy to boil itself through flashing al the way to 45 percent quality,
assuming atmospheric pressure.
To get from the 45 percent quality all the way up to 100 percent quality, you
need a heat input, and obviously that heat input is proportional to the size
of the leak rate. So the higher the leak rate, the higher the heat sync, the
more local cooling. The more local cooling you have, the more ability there
is for liquid to exist in that annulus, and it's the liquid environment which
is potentially corrosive to the low alloy steel.
The second point are the velocities, the magnitude of the velocities. For very
low leak rates, velocity, just a simple average mass balance velocity calculations
show very small velocities which are not consistent with erosion or potentially
flow accelerated corrosion mechanisms.
So,a gain, the leak rate is the controlling parameter in terms of the potential
for erosion or flow accelerated corrosion. So that's why we concentrate on varying
the leak rate.
Okay. Go to the next slide.
The leak rate also has another important determining characteristic, and that
is the leak rate determines the magnitude of deposits that will exit the pressure
boundary. As we've heard, of course, the concentration of boron in the primary
waters decreases over the fuel cycle from in the neighborhood of 2,000 ppm down
towards 100 ppm, in some cases lower than that, ten, five ppm at some plants
right at the end of the fuel cycle.
But if you integrate over the same time period for two different leak rates,
you'll get the amount of deposits being proportional to the leak rate.
The bottom line here from the analysis is that we integrate all of the results
together to determine the time frame for significant degradation and then correlate
the volume of wastage, material loss of the head versus the volume of deposits
produced, and, for example, at Davis-Besse it has been reported that there were
900 pounds of boron deposits on top of the head.
So we're trying to do analysis work in order to try to show how much wastage
you would expect as the amount of deposits on the head. Obviously hundreds of
pounds in deposits should be readily visible on top of the head. Much smaller
amounts of deposits may require the insulation to be removed.
All right. The material loss mechanisms. If we go to the next slide, we start
off on the corrosion or the chemical type of mechanisms. The first one here
--I'll just briefly touch on each one of these -- boric acid corrosion.
In the leak process, you can have a concentration occurring due to the boiling,
flashing and boiling, process which tends to concentrate the boron. So you can
end up with a concentrated boric acid solution.
However, if there's no oxygen, typically these sort of de-aerated boric acid
tests of low alloy steel show very low corrosion rates. So that's the first
thing to keep in mind.
The second potential mechanism here is deposits themselves. Could they be corrosive
without liquid?
And there have been some tests that have been attempted with some deposits on
top of low alloy steel and found to be very mildly corrosive in a human environment.
So that's the second potential mechanism.
Then we do have a crevice geometry here. We have the annulus. So potentially
there could be a crevice corrosion mechanism. Crevice corrosion is a mechanism
that's of concern in marine applications often. It's also a concern with the
waste packaging at Yucca Mountain.
So we've looked at crevice corrosion as a potentially significant mechanism.
We also have, as mentioned before, the low alloy steel is in contact with the
Alloy 600 nozzle. So there's a galvanic couple, and perhaps that could drive
a corrosion mechanism. Where that coupling, the low alloy steel will raise the
corrosion potential or the Alloy 600 will raise the corrosion potential of the
low alloy steel and provide the driving force for the corrosion. So we've also
looked at that.
Then the next mechanism coming down the list here is classic boric acid corrosion.
Now we have an aerated environment. There have been many tests performed in
this sort of environment. They're documented in the boric acid corrosion guide
book that's been published by EPRI , and you can have up to one to five inches
per year of corrosion shown in these tests where you have oxygen that's in the
solution.
Lastly here, molten boric acid corrosion. Boric acid deposits have a melting
temperature of about 340 Fahrenheit. So even without water, you can have a liquid
at the higher temperatures, and the question becomes: how corrosive is that
liquid? And so I'll have some comments on that molten salt type corrosion.
And this slide here are the flow type, velocity type mechanisms here, and the
first one being flow accelerated corrosion. That's a possibility depending on
whether or not there's a magnetite layer that may form on the low alloy steel.
This is, of course, a mechanism that is seen on the secondary plant in the piping.
So we've examined looking at the possibility of that having an influence on
the development of the process.
And then there are more just the straight erosion type mechanisms, flashing
induced erosion. If we think about gaskets that can develop leaks, you may have
a local region that may be a somewhat analogous situation here with erosion.
You hear the term "steam cutting erosion." That's just really another
term for flashing induced erosion. We have water droplets. So, therefore, the
term "droplet impingement erosion."
Single phase erosion of steam velocities as you boil water off all of the water
content in single phase steam and potentially you might have velocities of the
steam and potentially that could lead to a single phase erosion.
So that's an introduction to all of the mechanisms that we could come up with
for removing material.
This matrix here is a preliminary take based on the last two months of work
on how these mechanisms may stack up in terms of which ones are active. As I
mentioned, the first two have low rates. So we don't think they play a major
role in the progression.
Then we get to single phase erosion. We start with an initially tight annulus,
a gap on the order of 1/1000 of an inch radially there or perhaps tighter. So
initially if you have a leak, it may lead to velocities high enough to get erosion.
Now, once that annulus would open up, then the velocities would be reduced because
of the greater flow area. So perhaps for the initial tight annulus the single
phase erosion could be a factor or impingement erosion also.
I've got full accelerated corrosion listed here if the velocities are high enough.
Crevice corrosion. I can say that this is not a classic crevice corrosion type
system here because crevice corrosion is typically associated with materials
that passivate (phonetic), like stainless steels.
If we had -- crevice corrosion is driven by a chemical process where the anodic
corrosion reaction occurs deep down in the crevice, but the cathodic reaction
occurs away at the exposed surface on top of the head. If there was a liquid
film up on the top surface of the head, potentially you could have the driver
for a corrosion circuit from the outside to the inside deep down in the annulus.
However, in our case, if there's going to be a significant water film on the
outside of the head, in the top head surface, then we would expect there also
to be deposits in an acidic environment, which would lead to significant corrosion
rates themselves. So it would act as an anodic site up on the outside. So we
don't see this separation of the cathode and anode excites in the low alloy
steel due to the crevice corrosion, provided that you have the acidic environment
on the outside of the head.
But the next mechanism here, galvanic corrosion in the secluded type geometry
may be more of a possibility. We do have the coupling from the low alloy steel
to the Alloy 600, and that potentially does give you a driver for the corrosion.
However, there isn't enough data available in the literature to try to quantify
the magnitude of that mechanism. There just hasn't been a lot done with low
alloy steels and boric acid type environments with things to measure polarization
curves and so on. We haven't addressed that from a basic corrosion science standpoint
yet.
Molten boric acid corrosion here. I'm saying that it's possible, but we expect
lower rates. There isn't a lot of available data experimentally in terms of
trying to measure its corrosivity for low alloy steel. However, if we look at
the basic corrosion chemistry there, we know that the molten boric acid has
a lower -- the solubility of corrosion products are lower in molten boric acid
than in aqueous solutions. So that's one factor.
Electrical conductivities are likely to be lower in molten boric acid, and also
the oxygen and hydrogen ion concentrations are also likely to be lower in a
molten salt type solution.
So for some fundamental reasons we believe that the molten boric acid corrosion
is unlikely to produce the one to five inches per year that has been observed
with the aerated concentrated boric acid solutions, but it's still something
that has to be looked at.
Okay. So that takes care of this slide here. This next slide here really sums
up the analyses that have been done in terms of understanding the chemical environment,
looking at the pH through multi-Q calculations.
MS. WESTON: It's page 139 in the book.
MR. WHITE: And we've also performed thermal hydraulic calculations and heat
transfer calculations to try to quantify the temperature as a function of the
leak rate. We've calculated velocities as a function of leak rate, wall sheer
stresses, as I mentioned, the pH under various conditions.
So putting all of those things together, we've developed this degradation progression
here which really goes from the left side of the slide to the right side of
the slide as the leak rate may increase over time.
The top row of boxes here has a nozzle or weld condition. Early in time you
would just start out with a leak path to the annulus, but in a very small leak.
As that crack growth continues, that leak -- an axial through wall crack may
reach above the top of the weld for a significant distance. At Davis-Besse,
which would be associated with the far right area here, there was an axial through
wall crack that reached .9 inches above the top of the weld on the nozzle ID
and 1.2 inches above the top of the weld on the nozzle OD.
So there was a leak path that extended all the way through the nozzle a significant
distance above the top of the weld and leak rate calculations that we performed
as part of this work have shown that should result in a high leak rate, meaning
on the order of .1 gpm, which is consistent with all of the evidence for the
Davis-Besse nozzle number three.
So we have growing cracks, increasing leak rate as we go from left to right
across the page here.
MEMBER SHACK: Now, what does the pressure drop look like, say, with that .9
inch crack and I have a pressure drop across the crack into the annulus and
then I have the annulus -- the interference fit to the atmosphere? What's the
pressure drop across the crack and then across the interference fit?
MR. WHITE: Well, I do have some slides on that, but I don't want to go right
to them. What I would say is initially when you have that very tight initial
annulus of a mil, a half a mil or so, you may have also a significant pressure
drop in the annulus itself.
But as the annulus tends -- as you begin to have some material loss, very quickly
you'll reach a couple mils radial gap and the calculations show that you basically
have atmospheric pressure at a very large range of leak rates in that annulus.
So fairly early in the process we believe that we essentially have atmospheric
pressure in the annulus, and really the choke point in the flow is at the exit
of the crack.
And that's what I'm showing here on this line, is the annulus condition. Here
possibly hypothetically starting off clogged, but then opening up and allowing
more and more flow through, but it's really the crack that's more the governing
resistance to the flow.
Leak rates here. Well, we'll start over there. We have a hypothetical zero leak
rate. Contrary to experience, we had a nozzle with a leak path type crack, in
other words, a leak path reaching to the annulus, but there was no actual flow
making it to the outside to the top of the head. Then we would have a hypothetical
zero leak rate, and this column addresses that situation.
As we go to the right, we're increasing in leak rate, .001 gpm, .01 gpm as we
move to the right, and then up to the point greater than .1 gpm on the far right.
MEMBER ROSEN: Why do you say in your first column that you will at least have
some small amount extruded in that circumstance? This is the classic stealth
crack that we worried about.
MR. WHITE: Well, just thinking that if you're going to precipitate and go up
the annulus, you should be pushing out a small amount. I'm not claiming how
visible that's going to be.
MEMBER ROSEN: Pardon me?
MR. WHITE: I'm not claiming how visible that amount will be. I'm just saying
that you have a clogged up annulus with --
MEMBER ROSEN: It could stay subsurface you're saying. It says here at least
a small amount is extruded . Presumably you mean outside the crack in the annulus.
The extrusion results in deposits that are visible.
MR. WHITE: Right. Well, as I say --
MEMBER ROSEN: It's your chart. I'm just asking what you mean by that.
MR. WHITE: What the real experience has been over here in this chart, in this
column over here, we do have small amounts of deposits that come up that correspond
to small leak rates.
MEMBER ROSEN: I don't think I agree with you that it's been in the second column.
That column has a bottom line of seven pounds, and we've seen pictures where
there were very small amounts.
MR. WHITE: Less than seven pounds.
MEMBER ROSEN: Sure.
MR. WHITE: Yeah, much less, yeah.
MEMBER ROSEN: So the column on the left was what was operating in those conditions.
We had a lot less than seven pounds, and I'm trying to examine what happens
down at the end --
MR. WHITE: Right.
MEMBER ROSEN: -- the boundary of that.
MR. WHITE: So let me talk about if there is a zero leak rate what happens and
you don't have significant deposits that come out.
In that situation, in a hypothetical situation, we have no velocity. so you
have no erosion type mechanisms that could be active, and you would have no
cooling going on. So you would have a crevice environment there that's at 600
degrees approximately, the primary temperature.
But since this is a clogged annulus up to the point where the clogging is, you're
going to have pressurized water at the primary pressure. You're not having any
boiling going on because there's no flow. If there was boiling, there would
have to be a leak that would be actively going to the outside.
So there is no vaporization driven concentration mechanism with no flow at all,
and then as we heard earlier in John Hickling's talk, there's not going to be
oxygen down in that crevice environment. So there aren't the conditions that
would produce -- the corrosion rates would be limited to the low corrosion rate
s that had ben measured for de-aerated environments, and without a large concentrating
mechanism it should be even less than most of those tests which were done in
concentrated boric acid conditions.
MEMBER WALLIS: If I remember correctly, the analyses that were done to preclude
oxygen at the bottom of the annulus were done for fairly tight crevices and
straight fits that you have --
MR. WHITE: Right.
MEMBER WALLIS: -- in these designs. In order to get to the right of that diagram,
to get your temperature down, which you will need for the high corrosion rates,
does that preclusion of oxygen analysis still hold for the fairly wide annuli
that you're going to need to have the flurry?
MR. WHITE: No. Well, as we move all the way to the right here, aerated boric
acid corrosion once you have something that opens up to the problem, you definitely
would have the aerated boric acid corrosion as --
MEMBER WALLIS: Oh, okay.
MR. WHITE: The question becomes: at what point does the oxygen get down into
the crevice? It's obviously between those two points, and at this point we just
can't say exactly where that point is based on the work that's been done so
far.
MEMBER WALLIS: Okay.
MR. WHITE: It's just when you're very hot, the hot iron is going to be very
efficient at taking out the oxygen. It's when you have the cooling and the opening
up together and the higher velocities and the eddies that could form. Then you
could start to have oxygen coming down deep into the crevice.
MEMBER ROSEN: So to finish this discussion, the stealth cracking mechanism that
has been postulated that what we saw at Davis-Besse could be going on under
the surface, in your own words now, how likely is that here?
MR. WHITE: Well, the work shows that it's unlikely, taking in turn first the
case of no leak rate at all, having a pressurized annulus with single phase
liquid without a big driver for concentration and no oxygen. That would have
no active mechanisms.
As we move towards small leak rates, ten to the minus six gpm, ten to the minus
five gpm, for much of the cracking we see on the order of a cubic inch of deposits
that corresponds to a gallon of leakage in a year. That's two times ten to the
minus six gpm.
So as we approach ten to the minus five gpm, when you do the heat transfer calculations,
you don't get the cooling. So what's going to happen is that that annulus is
going to boil dry immediately right near the bottom of the crack, right near
the bottom of the annulus at the crack. So there isn't going to be liquid over
a significant volume or height inside that annulus. So that's really what's
preventing corrosion mechanisms that may potentially occur in the absence of
oxygen from really being significant.
I mean, this is just consistent with all of the experience out there for very
small leaks that show minimal material loss. You don't have the velocity mechanisms,
and you don't have very much liquid around at all. Perhaps it's all boiling
dry low in the annulus, and you need that liquid even to get something like
a galvanic type mechanism going.
CO-CHAIRMAN FORD: If I go from the left-hand side and approach the right-hand
side, you're having more and more conjoint requirements that are necessary to
get a Davis-Besse situation on the right-hand side.
MR. WHITE: Yeah, you're --
CO-CHAIRMAN FORD: And because there are so many conjoint requirements, annulus
size, exposed crack length, leak rate into the annulus. So you're precluding
the possibility of this being a generic phenomenon.
However, you don't have to be on the right-hand side to have a real bad situation.
Those EPRI and CE tests were one inch per year. So you've only got to get over
to the middle column before you've got potentially a fleet wide problem.
I use that obviously to make a point. It's not an isolated set of criteria that
you need. Am I overstating it?
MR. WHITE: If I were to draw this down here, these tasks that are on the upcoming
slides, I would probably draw this more towards this range. We can go over the
actual leak rates in these tests, but they were closer to the .01 to .1 gpm.
There was one test down at .002 gpm by Combustion Engineering that had a significantly
lower rate of corrosion than the other test at .01 and .1. So it's really the
.01 number that I'm taking from those tests as being sort of a critical value
based on that.
At Davis-Besse we believe that the leak was between .04 and .15 gpm based on
the unidentified leakage, based on the mass of deposits that were observed,
and other indicators. So that would put it all the way off to the right there.
MEMBER WALLIS: There seem to be various things I'd like to know more about.
I don't know the details of your analysis, but the way in which the annulus
clogs or doesn't clog or periodically extrudes whatever is in there, is that
just hypothetical?
One could postulate all kinds of things that could happen in an annulus in terms
of deposits and the way they can be pushed out or slowly slide out or do various
things.
MR. WHITE: Yeah, one possibility is that when the head cools you have the difference
in cold fission or thermal expansion. So the annulus tightens up, and as it
depressurizes, at that point the annulus comes back to that interference.
MEMBER WALLIS: It could slowly flow out although it's apparently solid. It could
slowly be extruded from the --
MR. WHITE: Right, being in molten form. What we're saying though here is we're
trying to show that regardless of those details, without having a liquid high
up in the annulus and without having any velocities to speak of, there are no
credible active mechanisms.
MEMBER WALLIS: Why is the velocity coming out of this hole zero feet a second
and not 1,000 feet a second?
MR. WHITE: Well, if you're postulating that the annulus is completely blocked
up.
MEMBER WALLIS: No, it's a crack. It's coming out of the crack. The crack tip
goes through, and the velocity -- it says liquid velocity exiting the crack.
It's coming out of the crack. So if we had a fairly broad crack and as it breaks
through, what is it going to say, a sonic flow at the exit poll? Why is it such
a low velocity coming out of the crack?
MR. WHITE: As we increase the leak rate, you're asking about the --
MEMBER WALLIS: Well, even at the beginning. I mean at any time why is it so
low? Why isn't it -- why couldn't it be much higher at the beginning?
MR. WHITE: Well, if we took the -- should we put up the slide?
MEMBER WALLIS: Maybe even do just a calculation of flow through very long, very
fine tube of a flashing liquid. It takes a pretty long tube before you stop
getting choking at the exit from the tube.
MR. WHITE: Let's show you what the --
MEMBER WALLIS: Maybe it's too complicated to get into now, but I'm surprised
that you couldn't get a much higher velocity under these circumstances.
MR. WHITE: Go to 544 in the other presentation.
MEMBER WALLIS: It depends a bit upon the shape of the crack. You do it as a
two-phase calculation of the flow in the crack?
MR. WHITE: Right. What we're really looking at here is we took as a flow area
the area opposite the crack.
MEMBER WALLIS: Well, if you're using Moody and Fauske and HEM, aren't those
models for choking?
MR. WHITE: No.
MEMBER WALLIS: Critical flow? Yeah, that's what Moody and Fauske deal with,
critical flow.
MR. WHITE: These are just slip models we're just using. We're just assuming
two-phased flow in a pipe, for example.
MEMBER WALLIS: Well, you're using a square root of density ratio.
MR. WHITE: Yes, right, right. Just to get a handle on the velocities, we were
interested in the velocities not right at the crack exit, but --
MEMBER WALLIS: You're basing it on the shape of the crack, not just on the --
MR. WHITE: Yes.
MEMBER WALLIS: -- flow rate.
MR. WHITE: I agree, I agree. As a first cut, we wanted to get some --
MEMBER WALLIS: So you assume something about the shape of the crack?
MR. WHITE: No. All we did was we took the flow area as the area opposite the
crack. As the flow turns, it's going to expand.
MEMBER WALLIS: Opposite the crack? No, it isn't. What matters is the flow and
the area in the crack itself.
MEMBER SHACK: This chart is showing a leak rate through the crack of a given
amount. this is the annulus velocity that you would get.
MR. WHITE: Right, as you're --
MEMBER WALLIS: It says here exiting crack. Maybe it's the words that are wrong.
MR. WHITE: I agree.
MEMBER WALLIS: If you were saying the velocity in the annulus -- I agree the
velocity in the annulus could be low, but the jet coming out of that crack could
conceivably be sonic, and that's going to do something in that annulus presumably.
MEMBER ROSEN: Graham.
MEMBER WALLIS: Yes.
MEMBER ROSEN: I have a slightly different model that at the crack itself, you
know, is a very labyrinth kind of thing, and it functions as a breakdown orifice.
MEMBER WALLIS: It's like a porous median, and then it maybe breaks through the
outside, a little hold.
MEMBER ROSEN: Just barely, and there's almost no -- the pressure drop through
this labyrinth and pathway is enough to --
MEMBER WALLIS: Well, maybe it is.
MEMBER ROSEN: -- initially create -- there's no velocity at all. I mean as it
first breaks through, it just drips.
MEMBER WALLIS: It certainly doesn't drip. It may come out with steam.
MEMBER ROSEN: Yeah, well, it flashes. I mean a little bit of liquid which is
completely broken down; pressure that's completely broken down in this labyrinth
--
MEMBER WALLIS: Well, that's your picture of it.
MEMBER ROSEN: -- drips out, drips and flashes.
MEMBER WALLIS: That's your picture of it.
MEMBER ROSEN: Yeah.
MEMBER WALLIS: I'd like to know what the reality is.
MEMBER ROSEN: Well, I'm just saying that there's a way to think about it that
creates very low velocity.
MEMBER WALLIS: Yeah, but there's also a way to think about it that gives you
1,000 feet a second or so.
MR. WHITE: We've done calculations to try to calculate the crack opening area.
So we can use those to give you some velocities also.
CO-CHAIRMAN FORD: If I could suggest we don't do that right now.
MR. WHITE: Okay.
(Laughter.)
CO-CHAIRMAN FORD: Time and what I want the committee to understand is where
they are in this overall approach. There are hundreds of questions, and you
get the idea.
MEMBER WALLIS: But then they're off by a factor of 10,000 in velocity, and it's
interesting to know.
CO-CHAIRMAN FORD: That's true.
On this diagram here, and it's the last question we'll take on this one.
MR. WHITE: Okay.
CO-CHAIRMAN FORD: It's my understanding the tech spec is one gallon per minute.
MR. WHITE: Right.
CO-CHAIRMAN FORD: And, therefore, all of the operating plants, if they could
detect, you could be right over the right-hand side there and have all of your
mechanisms which would be -- have one gallon per minute, absolutely okay. That's
the only criterion we're taking. That's correct, isn't it?
MR. WHITE: Well --
CO-CHAIRMAN FORD: I'm saying it could be done at -- according to the EPRI --
MR. WHITE: Right.
CO-CHAIRMAN FORD: -- tests of one inch per year, you could be down at .01 gallons.
MR. WHITE: The way that I look at the far right of the chart here is that the
calculations show if you have more than .1 gpm of low, you're likely to locally
cool all the way down to 212, the metal. So that you can have liquid that's
making it all the way out onto the top of the head. So there's going to be a
significant amount of boron deposits that are going to be wetted by that liquid
that's coming out, and it's going to be colder.
There are liquids that are going to exist over a certain area, and that is going
to be similar to situations that we're seeing at plants in the past that had
large leaks from up above sources that led to lots of deposits and wetting the
top of the head where up to about a half inch of material loss has been seen
in the past.
And some plants in Europe have also observed this with large leaks. So it may
not matter so much where the leak source is coming from one you have a large
leak here, that you can wet the top surface of the head, and you could have
corrosion possibly occurring from the top, from the top top.
CO-CHAIRMAN FORD: But my point is right this instant in time. Obviously there's
more work that has to be done, but right at this instant of time, if there's
any ability to your logic in that diagram --
MR. WHITE: Right.
CO-CHAIRMAN FORD: -- you'd better change your tech specs.
MR. WHITE: Well, what we're --
MS. KING: If you're depending upon leak detection only.
CO-CHAIRMAN FORD: Correct.
MS. KING: If you're looking at leak rate only, and what we're saying here is
we expect there to be significant visible evidence during a visual examination
of your head.
CO-CHAIRMAN FORD: Okay.
MR. WHITE: Yeah, it's important to put this in the context of a time frame.
CO-CHAIRMAN FORD: Right.
MR. WHITE: And based on the Davis-Besse root cause analysis report work, it's
believed that the high corrosion rates were occurring for four years, the last
four years, roughly an inch and a half of corrosion rate per year occurring.
But before that, it's believed that another four to six years, and of course,
it's not possible based on all of our information to nail down the exact time
progression, but it's believed there were another four to six years of leakage
that was occurring.
So it makes sense that as those cracks grew and you had more crack opening area
along that crack, that you would get the higher leak rate. So there still would
have been the four to six years to be able to detect something similar to or
larger than the amount of the deposits that were seen at other plants.
MEMBER SHACK: Just coming back to that then, in the four years you've got now
a one inch crack above the nozzle, and if you go back four years, how big is
the crack when you're getting significant of the low alloy steel?
MR. WHITE: Well, it's something that could be looked at.
MEMBER SHACK: Well, you know, presumably with this getting -- you know, as I
look at your mechanism, I keep coming up to some critical leak rate, which means
I need a critical axial crack size, which is far less than any structural limit,
and your argument would seem to tell me, you know, one inch minus four years
worth of crack growth.
MR. WHITE: Well, for this particular crack at nozzle number three, the evidence
indicates led to a leak rate on the order of .1 gpm. It was about an inch above
the top of the weld.
Typical growth rates argue for about a millimeter per year to perhaps up to
five millimeters per year, perhaps slightly higher based upon the French experiment.
MEMBER SHACK: Yeah, I would say a five millimeter crack.
MR. WHITE: Well, that would say that we went up to about 25 millimeters at five
millimeters per year, it would have taken five years to get to that point.
MEMBER SHACK: But significant attack then starts with the five millimeter through
wall crack is what you're arguing.
MR. RICCARDELLA: This is Pete Riccardella.
Bill, you know, that crack could have been there for a while though because
remember as the crack is growing, it's growing out of the residual stress field
and probably slowing down in that axial direction. I don't think a linear assumption
on crack growth is fair.
MR. HUNT: Let me just add one other point here. We do have a number of other
nozzles in other plants that have cracks just under one inch that seem to be
consistent with the lower leak rates over on the left-hand side of the chart,
and one of the things that we're looking at from a finite element standpoint
is, you know, where the transition occurs in this flow rate.
Steve Hunt, Dominion Engineering.
CO-CHAIRMAN FORD: Could I just in terms of managing time here? I know it's not
fair to you. Could you try and finish by quarter to two?
MR. WHITE: Sure.
CO-CHAIRMAN FORD: The other presentation also, the main purpose being to just
let everybody around this table know what the concerns are, what you're doing
to resolve those concerns.
CO-CHAIRMAN SIEBER: Let me just add one word about the applicability of leak
rate tech specs to this situation. If you look at the reactor coolant system,
there's a lot of places where it can leak a little bit, and that's through interconnecting
valves, through other systems, through safety valves, PRVs, pump seals and so
forth, and generally speaking leak rates like you're talking about on the head
are very small compared to some of these others.
This chart, which we haven't discussed yet is the one that shows weak rate versus
time. If you go back three cycles and you look at the leak rates in those early
-- the first two cycles, the leak rate is very low, which is pretty much typical
of PWRs, but it's probably enough to support the fact that you might have had
crevice leakage and annulus leakage of this nozzle.
So just to clarify that. That's not the only reason for the leak rate tech spec.
MS. KING: I think we discussed these.
MR. WHITE: Yeah, the next slides go over the basic -- this one might be worth
touching on here. Again, this is outlining the idea that was put forth by the
Davis-Besse root cause team as one possibility, that the material loss for nozzle
three occurred more from a top down type mechanism.
Perhaps these mechanisms like galvanic erosion led to some growth down deep
in the annulus, deeper than -- greater material loss deep in the annulus than
at the outside, but that once the leak rate reached that .1 gpm, then water
would reach all the way to the top, and it would have been a top-down type mechanism
where you have the aerated concentrated boric acid corrosion.
As the corrosion would have moved downward, then the surface area covered by
liquid would be less because now you would be going more into a pool geometry.
So this might explain the change in slope at the outside of the cavity.
In other words, the area at the outside of the head is greater than the area
as you move down.
And then also what might produce the shape of the cavity, the oblong shape,
it being longer in the downhill direction than the transverse direction?
Well, gravity would have displaced that pool, the initial pool on top of the
head in the downhill direction, and as you move down, that could explain the
shape.
CO-CHAIRMAN FORD: So you're still having no significant velocities in that pool?
MR. WHITE: Well, once things are opened up, velocities are not going to be that
high.
CO-CHAIRMAN FORD: So this isn't a solution mining type of thing where you go
down there and dissolve the rock with a jet of liquid? It looks like it. I mean
if you had a jet coming out of that crack, it would make a cavity something
like what was observed, I would think.
MR. WHITE: I'm not completely discounting the erosion type mechanisms. It is
a possibility, but we know, therefore, the aerated boric acid, concentrated
boric acid conditions, you can have the high corrosion rates.
So it seems consistent that that would have been the primary mechanism when
you got there.
CO-CHAIRMAN FORD: Okay. You're about to discuss those two tests. Just let me
make sure we know factually where we are right now. Right now you come up with
a series of hypotheses, qualitative hypotheses enunciating the conjoint requirements
to get the temperature in the annulus down to a lower value necessary to sustain
high crack growth rates, and you're relating that to leak rates from a practical
point of view.
Are there any tests planned for the near term to qualify that hypothesis?
MR. WHITE: Well, there have been discussions initiated between the MRP and NRC
Research as to what tests could be performed, and so we're in discussions with
the industry and with the NRC about--
CO-CHAIRMAN FORD: And how urgent are those?
MS. KING: We would expect this work to identify the appropriate tests to go
perform, and then we would take immediate action.
CO-CHAIRMAN FORD: Immediate being tomorrow.
MS. KING: Well, he needs to finish first.
CO-CHAIRMAN FORD: Well, I recognize that, but I still say it with some -- quickly.
MS. KING: Quickly, yes. I don't plan to wait until 2005. As soon as we can identify
what the appropriate tests would be, we would immediately start to pursue --
CO-CHAIRMAN FORD: And to fill hydraulic analyses. There are obviously a lot
of questions on thermal hydraulics in that crevice and how they change with
operating conditions, fit up and shrinkage, ovality (phonetic), and the dimensions
of annulus as they change in time. All of them will be addressed.
MR. WHITE: If we go to these slides here, we can go over briefly what's been
done in the past. This slide just touches on different types of boric acid corrosion
tests, but here we see some of what the mock-ups look like for testing that
was done sponsored by EPRI back in the '96-'97 time frame with different leak
rates simulating an annulus geometry with leakage, and if we could go to the
next slide, here's a specimen, one of the specimens from one of the six tests.
This was a leak rate of .01 gpm. The actual injection point is here along this
hole here. This is a thermal couple probe area here.
But the flow came through here, and then impacted on a stainless steel tube
that was inside this hole, and so you can see some of the corrosion that occurred
in this test and how it's deeper down in the annulus.
CO-CHAIRMAN FORD: The erosion was not on the impact point at this -- the back
side of the impact point.
MR. WHITE: Well, right. Here the flow is coming through --
MEMBER ROSEN: Remember the impact is on the stainless steel tube in this case.
CO-CHAIRMAN FORD: Oh, okay, okay.
MR. WHITE: Here's an example of one test that was performed in the past. Future
tests may want to look at the development of the corrosion rate versus time,
and there are techniques to try to capture that as a function of time.
They could try to quantify the environment carefully in terms of the temperature
along the leak path, in terms of the chemical composition along the leak path
and so on., the electrochemical potential.
So that's one area of potential testing. Other areas would go to the properties
of molten boric acid, its potential for being corrosive, for looking at the
galvanic mechanism.
In this test, one could decouple the stainless steel tube from the low alloy
steal, electrically isolate them and see if that was a major factor in terms
of the amount of corrosion that would indicate galvanic mechanism here.
This slide here just summarizes that those tests that were performed in the
'96-'97 time frame, along with tests that were performed in the late '80s by
Combustion Engineering of the pressurizer nozzle geometry, an inverted geometry
so that the nozzle was facing down rather than up out of the low alloy steel.
Those tests both produce similar corrosion rates, two to two and a half inches
per year, and one key thing form this testing was that for leak rates greater
than .01 gpm, as the leak rate was increased, actually the corrosion rate decreased,
and the belief is that that tends to indicate a corrosion type mechanism rather
than a flow erosion type mechanism and the reason is believed to be that the
higher flow rates would flush out the impurities, and you'd get actually a lower
boric acid concentration because you'd be with greater flow flushing out the
crevice.
These tests had interference or I should say gaps of about 5/1000 of an inch
radially, which is larger than the initial fit-ups for the CRDM nozzles, but
would be representative after that CRDM nozzle annulus opened up over some time.
MEMBER SHACK: I'm sorry. In the CE test, is that another one where the jet impacts
the tube or, no, the flow is coming --
MR. WHITE: That one they actually had a crack in a steam generator tube and
let the flow come from inside the steam generator tube and then go into the
annulus and then down, and that test, although it had a similar material loss
rate, the location of maximum corrosion was at the outside, at the exposed surface
down at the bottom.
MEMBER SHACK: But at least that one you did have a potential for erosion of
the low allow steel.
MR. WHITE: But the material loss didn't happen up there.
MEMBER WALLIS: Well, this two inches per year is for these particular tests.
MR. WHITE: Right.
MEMBER WALLIS: And I don't know that it's being predicted theoretically.
MR. WHITE: No.
MEMBER WALLIS: So there's no reason to suppose that two inches per year in this
test is the same as what you'd get in the reactor situation where flow rates
and commissions are not quite the same and the geometry isn't quite the same.
MR. WHITE: Also, I think an important factor is the amount of cooling that you
get because obviously it's difficult to mock up the way the reactor heats the
head with that large heat source. In these tests, there are cartridge heaters
that may be used to do the heating so that the amount -- the temperature drop,
the local temperature along the leak path could be much different.
MEMBER WALLIS: So you mean in the absence of a predictive method based on physics
and chemistry. I don't quite know what to do with two inches a year from these
tests. It could be ten inches a year or .2 inches a year in the reactor for
similar conditions because they're not going to scale it to the reactor condition,
unless I have some sort of a physical model.
MR. WHITE: Well, the objective of the test was to try to simulate typical conditions
as much as possible.
CO-CHAIRMAN FORD: I think the main point here, Graham, is that they've done
some preliminary hypothetical work and come up with a potential progression
of events, and there's an urgent need to do some confirmatory analyses and tests.
I think that's a fair --
MEMBER WALLIS: You're not going to be confirming anything. You're going to be
investigating and --
CO-CHAIRMAN FORD: Well, confirming the hypotheses and coming out with a prediction
of what happened in the actual plant.
MEMBER WALLIS: That's a long way to go.
MR. WHITE: But so far everything is consistent with the experience, I'd say.
The majority of the leaks we've had on the order of 35 leaking nozzles in the
U.S. The large majority of them have had small amounts of boric acid and no
measurable wastage or very small amounts of wastage.
That's consistent with not having a lot of liquid in the annulus with having
the temperature close to the primary temperature with the low velocities, and
we've had the one case, much larger in comparison, where the calculations showed
that liquid could make it all the way out into the top of the head, and that's
the case where we had the large corrosion.
MEMBER WALLIS: As soon as the inspection shows rivers running down the head
instead of just crystals coming out of the ground.
MR. WHITE: Right.
MEMBER WALLIS: That indicates there's liquid up there, doesn't it?
I've seen all sorts of photographs. I've seen these little popcorn around.
MR. WHITE: Right.
MEMBER WALLIS: And I've seen the popcorn with some sort of a river down below
it. That indicates that there's liquid. As soon as you see something flowing
down from the --
MS. KING: Well, but that doesn't necessarily mean -- in those photographs it
does not necessarily mean that the liquid -- the source was the PWSCC crack.
It potentially could have come from --
MEMBER WALLIS: What it means is that it was wet boric acid.
MR. WHITE: Also boric acid, you know, melts at 340 degrees, becomes molten.
It's very hydroscopic. So it like to pick up whatever moisture is in the air.
MEMBER WALLIS: It dries up as it --
MR. WHITE: So the appearance of deposits and the morphology of them is going
to change as the plant cools down.
CO-CHAIRMAN SIEBER: My impression was that the so-called lava flows were mostly
iron oxide and molten boric acid because it was hot enough that the liquid containment
pressure, that would vaporize right away, and you would end up with crystals
which would then melt and form these rivers.
That's at least my first impression of what I saw. Could you comment on whether
that was correct or not?
MS. KING: Well, I guess I could comment. The boric acid crystals initially early
on in the videos were brittle, and you could see that as they were being cleaned
off the head surface, and it was verbally reported later that they got --
CO-CHAIRMAN SIEBER: Chunks.
MS. KING: -- very hard and very difficult to remove, which would go towards
the molten boric acid.
CO-CHAIRMAN FORD: Could I suggest that we call a halt at this point?
MS. KING: Sure.
CO-CHAIRMAN FORD: Before we do that, could I ask just the staff? I know there's
a cracking action plan to deal with the cracking issues. Is there going to be
an associated degradation action plan?
They talked about in talking to you and talked about the appropriate tests that
they want to do to validate these hypotheses. Is there an NRR action plan associated
with that?
MR. WHITE: Not specifically at this time. We've asked Research to give some
idea of what we'd be looking at with respect to if we wanted to build a mock-up
and run a test or something like that, what it would conceptually cost us with
respect to dollars, et cetera.
I don't know if there's anybody here from Research. Bill, have you had a chance
to look into that at all?
MR. CULLEN: Bill Cullen, Office of Research.
The panel has asked a few times this morning about research that might be considered
going forward. Peter himself has specifically asked twice what's going to happen,
and just a few minutes ago Christine indicated that on the industry size there's
research that's being proposed.
I have received, as you can imagine in the last three months all kinds of proposals
from all kinds of people who have considered themselves to be experts in boric
acid corrosion. I'm currently going through those things trying to figure out
what's good, what's bad, and what would be helpful.
Additionally, as Bill Bateman has just indicated, NRR themselves has asked Research
to come up with a plan.
So combining all of those requests and all of that input, I am I would say nearing
the end of the line on deciding what it is that we're going to do. Even as we
speak there's some proposed funding documents that are circulating in the next
building. So we will, I think, know within a very short time how much funding
could be made available for Office of Research sponsored funding to look into
some of these issues dealing not only with the corrosion aspects of this, but
also some nondestructive inspection programs that might, you know, when they
get implemented within the industry, serve to help find this sort of situation
or this sort of degradation long before it gets as far as it did get in the
Davis-Besse.
Also have some other plans about I would like to maybe do some sensor development
because I think we've talked about the right element monitors and the containment
air coolers being credit up (phonetic), going along that line or down that road.
I think there's some sensor developments, some instrumentation development that
could be undertaken that would also help.
CO-CHAIRMAN FORD: Will that be covered in part by Ed? Are you going to be talking
later on on the inspection?
MR. CULLEN: It's not my position to say, but Ed Hackett will not be presenting
today. Mark Kirk will be presenting a little later on, PFM.
MEMBER BONACA: Since we are on the subject of inspection, just one thing that
is a no brainer, doesn't need the research. Two things actually.
One, are we going to make some criteria on what the licensee has to do when
he finds he has flanges leaking?
I mean one of the problems at Davis-Besse is that they manage the leaks. They
only fix some, and then they decided to put off until the next outage some.
On deposits the same thing happened. They simply had a certain time allotted
for removing deposits. They removed what they could at the time, and then they
just started again without removing all deposits.
Are we going to establish some criteria for this? It's not only inspections.
It's what we're going to do with what we find after we inspect.
It seems to me that as a minimum one would expect that if you find leakage you
fix the flanges before you restate. That's become a priority.
MR. BATEMAN: Yeah.
MEMBER BONACA: Also it would be the removal of boric acid deposits.
MR. BATEMAN: Yes, you're right. If licensees do find leakage at their flanges,
they do repair the leakage before they restate.
We're going to talk a little bit later this afternoon on the inspection plan
methods and frequency. So I think that will probably address the other --
MEMBER BONACA: Because, you know, those things, I mean, are no brainers. You
don't need research for that. If you do that, you're going to find where the
problem is. You know, you're not going to be stumbling and propagate the problem,
cascade from cycle to cycle as it happened at Davis-Besse.
I would expect that that would be a requirement that it would be very reasonable,
in fact.
MR. BATEMAN: Yeah.
MEMBER BONACA: If you had accumulation of pounds of boric acid crystals on top
of the head, I think the requirement should be remove them all. You don't restart
until you've done that.
MR. BATEMAN: Yeah.
MEMBER BONACA: I think any licensee who can think with his own head in this
situation would do that.
MR. BATEMAN: From the results of Bulletin 2002-01 inspections, I think we feel
at this point that Davis-Besse was an anomaly, and I think maybe we have a tendency
to try and apply that to everybody else, but we haven't discovered anybody else
in the industry who has come anywhere near close to having the boric acid accumulations
on their head that Davis-Besse had.
MEMBER BONACA: But you understand the requirements I'm discussing here are reasonable
actions.
MR. BATEMAN: Absolutely, absolutely.
MEMBER BONACA: And those could have prevented so much of this pain and attempt
on our part now to try to foresee the future. It's going to be very hard to
do anyway, but I think fundamentally some basic ground rules, and I would call
them almost housekeeping for a plant.
MR. BATEMAN: I think it's basically more of an implementation issue. Licenses
have a boric acid inspection corrosion program based on their response to generic
letter 8805, and we found with the exception of one licensee that they're implementing
it.
I think it was more of an implementation problem as opposed to a program problem.
I think the programs are out there. It's how well are they implemented.
CO-CHAIRMAN FORD: Could I ask that we move on at this stage?
And, Larry, could I ask the two of you to do both of your presentations by quarter
to, finish them by quarter to three, at which point we'll take a break?
MR. MATHEWS: I really don't have much to say. We could skip the collateral --
it's just two slides.
CO-CHAIRMAN FORD: All right.
MR. MATHEWS: We really haven't done anything since last time.
CO-CHAIRMAN FORD: Okay.
MR. MATHEWS: We can come back later and talk about it.
CO-CHAIRMAN FORD: If the rest of the committee -- if it's okay with them, we
will skip Larry Mathews on collateral damage. There's not a lot that has been
done since then, since the last time we met.
Pete.
MR. RICCARDELLA: Okay. I'm Pete Riccardella from Structural Integrity Associates.
We were contracted back about September of last year to develop a probablistic
fraction mechanic's model for this top head degradation and cracking issue,
and the focus of that model is primarily looking at probabilities of the growth
of large circumferential cracks and nozzle ejection.
What I plan to present today is somewhat of an overview --
MEMBER APOSTOLAKIS: Excuse me. Where are these slides?
MS. KING: These slides follow the crack growth rate.
MS. WESTON: Page 46 on this handout.
MS. KING: There you.
MS. WESTON: And they start on page 59 in the book.
MR. RICCARDELLA: Are we set?
My purpose today is to present an overview of this model, and I should say that
in the period of time since September, we a have had several meetings with the
NRC staff. There's been several interactions, both teleconferences and meetings
where we've discussed, traded ideas on the methodology.
And in fact, we've gone back and made some modifications, adaptations to the
model based on NRC staff input.
So first I'll give an overview of the methodology and then talk about some PFM
analyses that we've performed in support of the propose MRP inspection plan.
The key elements of our probablistic fracture mechanics model are listed on
this slide. We have an experiential based probability of leakage model. We don't
try to model the initial nucleation and growth of the crack. We're basically
just going back and looking at based on experience probability of leakage versus
time, and then we take it from that point.
We have a fracture mechanics model for stress intensity factor in which we've
considered both part through wall and through wall cracks in different nozzles,
different locations on the head, different places on the hillside.
But the assumption in our fracture mechanics modeling is, I believe, a conservative
one, is that once we detect a leak, we assume that we instantaneously have an
axial crack which has branched and turned to a circumferential crack and is
already 30 degrees of the circumference.
So that's the starting point for our analysis. We've compared that to looking
at a leak and then getting multiple initiations, reinitiating new cracks around
the periphery, and we believe that our model instantaneously assuming a 30 degree
circumferential crack is both conservative and less arbitrary than you get by
trying to model these multiple reinitiations of circ. cracks.
It also agrees at least with the anecdotal evidence of how the circ. cracks
developed at the Oconee plant, that they tended to be more like axial cracks
that branch into circ. cracks rather than reinitiated circ. cracks.
MEMBER APOSTOLAKIS: Where does the work that was presented earlier on the rate
of growth of cracks fit into this?
MR. RICCARDELLA: We have the next slide a key. A key element of our model is
the statistics of crack growth, and I'm going to show you how I've used the
work that John --
MEMBER APOSTOLAKIS: I'm a little confused by the first bullet there.
MR. RICCARDELLA: Okay. Well --
MEMBER APOSTOLAKIS: What does it mean, that you're already having a leakage?
MR. RICCARDELLA: Well, what we're assuming is that at a certain period of time
we have a leak.
MEMBER APOSTOLAKIS: Right.
MR. RICCARDELLA: And then when we have a leak, we assume that at that point
in time, we have an axial crack that branches into a 30 degree of circumference
circ. crack, and then we use the crack growth to analyze the progression of
that 30 degree of circumference crack out to failure, out to 300 or 330 degrees,
whatever it is that produces ejection of the nozzle. Okay?
So we are arbitrarily giving up that initial nucleation and growth portion of
it.
MEMBER APOSTOLAKIS: Which could have included the crack growth rate again, right?
MR. RICCARDELLA: Some amount of crack growth to get to 30 degrees, but we're
kind of giving that away.
MEMBER APOSTOLAKIS: All right.
MR. RICCARDELLA: And we're saying we start there basically.
MEMBER SHACK: Well, you start there, but you take the Weibull model into account.
MR. RICCARDELLA: Yeah.
MEMBER SHACK: So I mean, you accounted for it in a different way.
MR. RICCARDELLA: Exactly.
And then finally in our model we have the ability to look at the effect of inspections
on this cracking and on the probability of an ejection. We can look at different
inspection intervals, as well as different levels of reliability for different
types of inspections, and we have some assumptions we've made regarding the
probability of detection.
If you have a leak and you do an inspection, what's the probability of detecting
that? Also the probability of detection for ultrasonic or other --
MEMBER APOSTOLAKIS: So where does that go?
MR. RICCARDELLA: Huh?
MEMBER APOSTOLAKIS: The probability of detection?
MR. RICCARDELLA: Down here, effective inspections.
MEMBER APOSTOLAKIS: And you also have a model for the probability of detecting
and doing nothing about it?
MR. RICCARDELLA: No, no. The assumption is that --
MS. KING: Do you mean no inspection?
MEMBER APOSTOLAKIS: I know it is there --
MR. RICCARDELLA: -- if it's detected you fix it.
MEMBER APOSTOLAKIS: -- but I don't have time to do anything. I'll work as best
as I can.
MS. KING: Essentially I think what you're saying is the effect of not completing,
not completing it.
MEMBER APOSTOLAKIS: Not doing anything about it.
MS. KING: Yes. We can take this model and do no inspections, and you can see
what the --
MR. RICCARDELLA: Well, no, not fix it. He's saying you detect it and you find
it and you don't do anything.
MS. KING: Oh.
MR. RICCARDELLA: We can conservative -- let's just say we'll conservatively
bound that in our probability of detection because we're using some pretty low
numbers for probability of detection.
CO-CHAIRMAN SIEBER: Let me ask a question about that a little bit. I've been
thinking about it since I'm the ultimate determinist. If you find a crack in
the through wall or greater than 40 percent and it's not at a mechanical joint,
okay, you know, a bolted joint, the code applies to that, does it not? And the
code says you've got to repair it.
MR. RICCARDELLA: Oh, yes.
CO-CHAIRMAN SIEBER: Otherwise you're in violation of the boiler and pressure
vessel code; is that correct?
MR. RICCARDELLA: Yes.
CO-CHAIRMAN SIEBER: And so everything you find that is greater than 40 percent
has to be repaired, right? Is that true?
MR. RICCARDELLA: I think it's 75 percent.
MS. KING: Yes.
MR. RICCARDELLA: It's more like 75 percent than 40.
MEMBER SHACK: I'm not sure there's an exact code section that applies.
CO-CHAIRMAN SIEBER: Well, you can calculate how much margin you have.
MEMBER SHACK: Yeah, and in fact, there was a memo from Jack Strosnider to the
MRP that says, "Here's an acceptable set of acceptance criteria until we
figure out what's the right thing to do."
CO-CHAIRMAN SIEBER: But a through wall crack you have to repair.
MEMBER SHACK: Yeah, if it's leaking, yeah.
MS. KING: Actually the flaw acceptance criteria is related to cracks that intersect
the pressure boundary at specific depths and the location of that crack.
MEMBER LEITCH: Ninety-five percent limit after the next operating period.
MS. KING: Right.
MR. RICCARDELLA: But the assumption in a probablistic model is that if you inspect
and find a crack, you fix it. We take it out of the population as far as possibly
proceeding to a nozzle ejection.
MEMBER APOSTOLAKIS: But, I mean, how real is that? I mean, why are we doing
it? Is somebody else dealing with the issue of if you find it, you decide to
do nothing about it, you know?
MR. RICCARDELLA: No. I think the inspection plan and the code tell you what
you have to do if you find it.
MEMBER APOSTOLAKIS: But if the code is implemented correctly, I expect the probabilities
to be very low. That doesn't tell me anything. I mean, the finding is the boric
acid corrosion control program at the site included both cleaning and inspection
requirements, but it was not effectively implemented.
Now, to tell me that, you know, I believe that they will find it and do something
about it doesn't address this issue.
CO-CHAIRMAN SIEBER: Well, one of the problems --
MEMBER APOSTOLAKIS: Now, that's not fracture mechanics, but that's the issue.
CO-CHAIRMAN SIEBER: One of the problems with visual inspection is you're beyond
70 percent. You're through wall, and then if you have, for example, the CRD
in flange, which I think is a bolted joint, right? We had welded joints, but
you know, those by code can leak. Okay?
So the issue is can you do a visual inspection with all of this boric acid that
leaked down laying on top of the head, okay, and if it's there, what do you
do about it?
I think that's part of the issue, which again tells me that sooner or later
you have to go to a volumetric kind of inspection to be able to satisfy the
requirements of the code.
MR. RICCARDELLA: Yeah, I think we might be able to address the question a little
better if I get a little further into the presentation and talk about exactly
how we're using the probablistic fraction.
MEMBER APOSTOLAKIS: Now, on this slide it says theta. What is the expression
for the Weibull?
MR. RICCARDELLA: What is the expression?
MEMBER APOSTOLAKIS: Yeah. Do you have the mathematics of it so that I know what
theta is?
MR. RICCARDELLA: Give me the one with the curve. I have an actual curve of the
Weibull. Okay? This Weibull paper, it's a standard, two parameter. I'm sorry
I can't quote it off the top of my head. Perhaps Glenn can bail me out and give
me the expression.
It's a standard two parameter Weibull.
MEMBER APOSTOLAKIS: Yeah, but it has several different expressions.
MEMBER SHACK: Theta is like the mean value.
MEMBER APOSTOLAKIS: Well, but it's not.
MEMBER SHACK: It's not, but it's like.
MR. RICCARDELLA: Theta is like the 63 percent cracking, and there's a function
of service. Now let's go back to the previous slide.
We have a Weibull analysis that was actually developed by Dominion Engineering
based on all of the B&W plants and the cracking that's been experienced
in those, and we made the assumption of a Weibull slope of three.
Actually, Christine, if I could go to the next slide, if you look at the actual
data, you would actually predict if you just did a pure analysis of the data,
you would predict a much steeper Weibull slope, like about a Weibull slope of
nine. That's the curve that --
MEMBER WALLIS: Once they get to 20 they're-- oh.
MR. RICCARDELLA: Pardon me?
MEMBER WALLIS: Once they get to 20 it's --
MR. RICCARDELLA: Pretty much, yeah, but we believe that there is something else
going on here, that there is somewhat of an inspection transient going on and
that some of these were leaking earlier in time, but we didn't start doing inspections
until pretty late.
And so --
MEMBER SHACK: Except the Oconee 3.
MEMBER WALLIS: Still it's cumulative.
MR. RICCARDELLA: That's true, that's true. Well, except if I use a steeper slope,
I'd get a less conservative result. So we're using the slope of three.
MEMBER SHACK: Steep slope is sort of good news. So it's good news and bad news,
but your header, you really meant the slope is three, not theta.
MR. RICCARDELLA: Yes, you're right.
MEMBER SHACK: You've totally confused us here.
MEMBER WALLIS: Weibull analysis is a prediction of these lines. Is that what
it is?
MEMBER APOSTOLAKIS: They assume the functional form. That's what it means, Weibull
analysis.
MR. RICCARDELLA: Yeah, well, this is actually a Weibull paper. So, you know,
the shape of the equation is built into it.
MEMBER APOSTOLAKIS: How many parameters?
MR. RICCARDELLA: Two parameter Weibull.
MEMBER APOSTOLAKIS: Two parameter. And then by fixing the slope, essentially
you end up with one parameter.
MR. RICCARDELLA: One parameter. That's right.
MEMBER APOSTOLAKIS: Now, can you tell me what the cumulative fraction of number
of leaking CRDM nozzles is? What does that mean?
MEMBER WALLIS: Just what's the fraction of the number of the leak. If there
are three out of 100 or --
MEMBER APOSTOLAKIS: So if I have 150 of them, this will tell me the fraction
of them that are leaking?
MR. RICCARDELLA: Yeah, Oconee 3 had this fraction leaking, and then the next
inspection they had that fraction leaking in that plant.
MEMBER APOSTOLAKIS: So this is between inspections?
MR. RICCARDELLA: Between inspections? There's only one plant on here that's
inspected twice, and this is the time period between inspections, from here
to here. The others are just until the first inspection. That's the fraction
of nozzles that were found. So there --
MEMBER SHACK: With two inspections he can calculate both of the parameters.
With one inspection, he has to assume one.
MEMBER APOSTOLAKIS: So let me understand what you just said. The other plants
are not inspecting at all?
MR. RICCARDELLA: No. This is just -- this Weibull analysis is just based on
the B&W type plants.
MEMBER APOSTOLAKIS: Right.
MR. RICCARDELLA: Because they've had seven out of seven leakers. The others
that have been inspected, most of them have had non-leakers. There have only
been two other plants that had leakers, and they're not shown on here, but they
fit very well into this group of data. That Weibull fit fits all nine of the
plants that have leaked, that have had leaks fairly well.
And, in fact, at the time this chart was produced, the Davis-Besse hadn't been
inspected yet, and the mean prediction was here, and when we actually did the
inspection, they came out -- the point falls very, very close to that.
MEMBER APOSTOLAKIS: But, I mean, this is the result of some inspection scheme,
isn't it? So what was that inspection scheme? How often do they inspect these
things?
I'm trying to understand that.
MS. KING: This data came from the Bulletin 0101 inspections.
MEMBER APOSTOLAKIS: Yeah, and? And? Don't assume that I know.
MR. RICCARDELLA: It was just the first inspection of a large number of plants,
and what we did was, you know, based on a lot of data from stress corrosion
cracking behavior, this type of material, we concluded that the largest slope
that we'd expect to see is three, and so we fit the data with a slope of three.
As I said, if you did a pure fit of the data to solve for both slope and theta,
you'd end up with a much steeper slope.
MEMBER WALLIS: You don't put in any data at all really.
MR. RICCARDELLA: Pardon me?
MEMBER WALLIS: You're just drawing some lines. You're not fitting any data.
MR. RICCARDELLA: No, not really. No, we're just saying where does the slope
of three best fit between that group, that group of data.
And then what we did is we had an upper bound and a lower bound, and in our
Monte Carlo modeling we assumed a mean and then a variation about that mean,
and you know, we assume a Weibull slope if you go back to the previous --
MEMBER APOSTOLAKIS: So these are --
MR. RICCARDELLA: -- 15 with a -- nine is the worst case and a 21 is the best
case.
MEMBER APOSTOLAKIS: And these are the fifth and 95th percentiles? In the Monte
Carlo simulation, how will these --
MR. RICCARDELLA: They're triangular actually. We're using a triangular distribution
for this particular --
MEMBER APOSTOLAKIS: So these are 100 percent?
MR. RICCARDELLA: Yeah, 100 percent and zero percent.
MEMBER SHACK: Now, Peter Scott with a much larger database to work with comes
out with a 1.5 slope, which is in your case more conservative. So your three
isn't conservative.
MEMBER APOSTOLAKIS: Where was this information?
MEMBER SHACK: When you do this kind of analysis for the French plants where
you actually have a larger set of data so that you don't have to assume the
slope, you get a number of 1.5 instead of three.
MEMBER APOSTOLAKIS: So the curve then -- the straight line would be very almost
horizontal.
MEMBER SHACK: Well, much closer to --
MR. RICCARDELLA: It would be shallower like that, yeah.
MEMBER SHACK: Which is bad because you get earlier initiation.
MR. WHITE: Bill, can I address that?
Dominion Engineering did the work of determining that three was the appropriate
slope to use. One major source is MRP Report 66, which just came out earlier
this year. In that work, the investigators looked at a large set of available
data mostly for crack initiation, and the best fit Weibull slope to that large
set of data, and I can't remember exactly the number of data points, but this
was a much larger set, I believe, than Peter Scott was working with.
And the best fit was 2.7 for the slope.
MEMBER SHACK: Was that steam generator tubes or nozzles?
MR. WHITE: It was on all available crack initiation tests for Alloy 600.
MEMBER SHACK: Okay. Because the earlier results from Dominion just looking at
steam generator tubes gave numbers much closer to the 1.5.
MR. WHITE: Well, we've also --
MEMBER SHACK: Original Gorman, you know, reports.
MR. WHITE: One of the presentations that I made at the meeting on May 22nd with
the NRC went over this in a little more detail, but we looked at the available
data, and there is a range that's observed, but if you look at that, there are
for some steam generator field experience in some locations in the tubes higher
PWSCC slopes than others, and three seems to be appropriate based on that also.
MEMBER SHACK: Well, in the steam generator, it's always conservative to take
the higher slope because that's predicting lots of tubes to fail.
MR. WHITE: I'm just talking about actual observed ranges of slopes for role
transition PWSCC for various locations, U bands, for example, and I can show
you that data if you'd like.
MEMBER SHACK: And compare with my data.
MR. RICCARDELLA: I think we could do the analysis with a slope of 1.5. I don't
think it would have a huge effect on the probablistic fraction mechanics results.
I've done it for nine and three, and I've found that that was worth maybe about
a factor of two. You saw that.
And I think the difference between three and one and a half would be even less
of an effect than that.
And considering that I benchmarked it against Oconee, it really wouldn't make
that much of a different. Okay?
MEMBER APOSTOLAKIS: Now, you're going to use this for future?
MR. RICCARDELLA: Yeah, we
re using this in our Monte Carlo analysis to create --
MEMBER APOSTOLAKIS: Right, but the inspections will have some sort of a period
in the future?
MR. RICCARDELLA: Yeah, we can assume various inspection intervals and --
MEMBER APOSTOLAKIS: So why does this apply?
MR. RICCARDELLA: Well, under the assumption of no inspection. We have a time
based Monte Carlo analysis that we start at zero, and we say at so many years
we would predict this many leak, and a couple of years later we would predict
this many leaks. So we can do a complete analysis with no inspections, and then
we can come back and superimpose inspections on top of that and see what the
impact is on those results of different inspection intervals.
Okay. So that's just the starting point. So here is, for example, for a 600
degree plant, beta equal to three, theta 15 plus or minus six. These are the
assumed times. So if we're at ten years and we're just the mean case in this
particular case, that would be about a 25 percent probability of leakage. If
we're out at 16 years, that's a 60 percent probability of leakage, but we vary
it between these extremes in the Monte Carlo modeling
So that's the very starting point of the analysis.
MEMBER APOSTOLAKIS: The triangular distribution you mentioned.
MR. RICCARDELLA: Yes, it's between these extremes.
MEMBER APOSTOLAKIS: In the vertical sense or the horizontal sense?
MR. RICCARDELLA: Yes, in the vertical sense.
MEMBER APOSTOLAKIS: In other words, ten effective years has a probability anywhere
between .1 and .7.
MR. RICCARDELLA: Yes, and incidentally, this is the probability of first leak
in a head of a certain number of nozzles. In other words, it's not the probability
of a leaking nozzle for any individual nozzle. It's the binomial probability
given 69 nozzles that you'll have at least one leak.
MEMBER APOSTOLAKIS: Oh, so this probability on the left is not the previous
probability?
MR. RICCARDELLA: Well --
MEMBER APOSTOLAKIS: The previous was accumulative.
MR. RICCARDELLA: Yeah.
MEMBER APOSTOLAKIS: Now it's something else.
MR. RICCARDELLA: The direct relationship, if you go back --
MEMBER APOSTOLAKIS: Why is that still viable? It's not viable anymore, is it?
MR. RICCARDELLA: Yes, it is. In fact, it is, and the relationship is theta for
a leak in a nozzle is equal to -- I'm sorry. Theta for the first leak is equal
to theta for a leaking nozzle divided by the beta root of N, where N is the
number of nozzles. There's a direct relationship between theta. The slope stays
the same, and there's a direct relationship between theta for that given leak.
MEMBER APOSTOLAKIS: Do you have that derivation someplace?
MR. RICCARDELLA: Yeah. Yes, we do.
MEMBER APOSTOLAKIS: And we can look at it?
MR. RICCARDELLA: You certainly can. I have it on my laptop.
MEMBER APOSTOLAKIS: Mag, can you please get that document so we can look at
ti?
MS. WESTON: I'm sorry. What is it you want, George?
MEMBER APOSTOLAKIS: The document.
MR. RICCARDELLA: Derivation of the relationship between theta for first leak
and theta for nozzle leakage in general.
MS. WESTON: And where do I find it?
MS. KING: On his laptop.
MR. RICCARDELLA: We'll give it to you.
MS. WESTON: Is it in your original presentation that you did to the staff?
MR. RICCARDELLA: No.
MS. KING: No, we haven't had this question yet.
MEMBER APOSTOLAKIS: This is the first question that you get that you haven't
had before?
MS. KING: For the derivation. No one has asked to see the derivation yet.
MR. RICCARDELLA: You know, that just gets us to the point where we assume a
leaking nozzle. Then we have to say, "Okay. How do we evaluate a nozzle
growing from this assumed condition at leakage, which is the 30 degree crack,
to a large circ. crack that could potentially lead to ejection?"
And so we've developed a series of finite element models with cracks of different
sizes and different depths. This is the model we use for a through wall crack,
180 degrees, and this is a crack that is assumed to initiate on the up hill
side of a hillside penetration.
MEMBER APOSTOLAKIS: Again, I'm a little slow here. You said that basically you
use a binomial distribution there to get with the expression for the first leak.
MR. RICCARDELLA: Yes.
MEMBER APOSTOLAKIS: Now, what happened at Davis-Besse, as I recall, was that
there were three nozzles that were adjacent. Now, in the binomial, of course,
you assume independence. So is that a valid assumption?
MR. RICCARDELLA: Yeah, if we look at the distribution over all the plants that
have had leaks, it is pretty random distributed around the head.
MEMBER APOSTOLAKIS: No, but when --
MR. RICCARDELLA: Davis-Besse happened to have three, but those happen to be
three that were out of the same heat of material that happened to be a particularly
susceptible heat of material.
MEMBER APOSTOLAKIS: Right.
MR. RICCARDELLA: Okay? But we believe that as far as time to leakage, there's
no geometric dependance at any of the nozzles at any particular location in
the head. The Oconee nozzles, the ones that leaked, tended to be toward the
periphery. The Davis-Besse nozzles tended to be near the top dead center.
Now, in terms of the tendency to develop large circ. cracks, we believe there
is a dependence on where you are on the head, but the time to leakage, we believe,
is pretty independent.
MEMBER APOSTOLAKIS: Okay.
MR. RICCARDELLA: So let's go back to that one. This is the model we use for
the through wall crack initiating here, running parallel to the weld. The red
Xes that you see in the bottom of the condition we applied for the J-groove
weld, and here's the crack tip. You see the added refinement that we put in
the mesh of the crack tip.
And this model enables us to calculate the stress intensity factor at that crack
tip, and we've run this for 30 degrees up through 330 degrees, and we have K
versus crack size.
We've run it for nozzles of different angles, you know, top dead center going
all the way out to the hillside like this, and we also use gap elements on the
back side of the crack to represent the constraint provided by the vessel wall.
This is where it shrunk fit into the vessel.
Next.
We also have part through wall crack model where we consider an axial crack
that is branching and turning into a circ. crack, and this is what we use for
the shallower crack configurations to calculate the K.
So this is what we use to calculate the stress intensity factor K that we use
in conjunction with the crack growth expressions that John Hickling presented.
You have to get a stress intensity framework, and John said, well, it's typically
25 to 30 or 35. These are the models that we use to develop that stress --
MEMBER WALLIS: How independent is this model that you choose? I mean you could
have chosen really different geometry, couldn't you?
MR. RICCARDELLA: Yes. Could I have the next slide?
What we've done, we've assumed a -- for most of the analysis I'm going to present
now and really all that we have done right now is the B&W type plant. Okay?
So we specified that type of geometry, and then we've looked at nozzles of zero,
18, 28, and 38 degree angles, and so that takes us from top dead center to the
most hillside.
And in the Monte Carlo analysis, we bend the nozzles into one of these four
categories. You have a nozzle for that and like every single category, and what
we find -- and then we've also looked at cracks emanating from the uphill side,
growing down, and then also emanating from the downhill side and growing up.
And for this particular geometry, what we found is that the uphill side is much
worse and also that the stress intensity factors get higher particularly for
the longer cracks as you get further away from the center.
MEMBER WALLIS: They depend upon what you are assuming about the shape of that?
MR. RICCARDELLA: No -- well, yes, they do to some extent. They depend upon that.
We've made what I believe is a conservative assumption on the shape of the crack.
They also depend on the residual stress, which is the size of the weld and the
welding parameters, and we currently have underway analyses of a CE type head
and of a Westinghouse type head so far.
MEMBER WALLIS: So with these assumptions you have to make about the crack shape
and all of that stuff, what's the uncertainty in these Ks?
MR. RICCARDELLA: I would say it could be as much as a factor of two.
MEMBER WALLIS: A factor of two. So that covers pretty well the range of the
data that we were looking at this morning.
MR. RICCARDELLA: No, but you'll see that when we get into the Monte Carlo modeling
that the effect of the uncertainty in crack growth rate is like factors of ten
to 20, and they tend to overwhelm. You know, the scatter in that lot normal
or distribution overwhelms, and remember we're using --
PARTICIPANT: The variation in --
MR. RICCARDELLA: Yes, the variability. And remember there's an exponent of one.
So it's pretty much a one-to-one relationship between K and crack growth rate.
So I think that, you know, further sharpening the pencil on stress intensity
factors isn't going to make that big an effect.
MEMBER WALLIS: -- aware of how uncertain it is.
MR. RICCARDELLA: Yeah, I would say it could be as much as a factor --
MEMBER WALLIS: Well, when you give us 26.9, it's probably anything between 20
and 35 or something like that.
MR. RICCARDELLA: We have analysis -- the highest number that we have anywhere
here is this 38. We have an analysis for another plant where that's as high
as 60, okay, for a different plant type, different residual stress.
And then we went ahead and did the probablistic fracture mechanics on that,
and it had maybe a factor of two influence on the probability of nozzle ejection.
MEMBER SHACK: Zero angle nozzles. So I mean this is a cylinder under pressure.
Pi squared over pi R squared P. Why am I not getting at least a pressure K that's
going up by the time I'm getting the 300 degrees?
MR. RICCARDELLA: I'm not sure. That's something I've got to go back and check
into. I think it has to do with the distribution of the street. You know, it's
a through wall where you've got residual -- it's residual plus.
MEMBER SHACK: I've even got a large gap here. So I've really got bending at
this point, right? You're letting this sucker bend.
MR. RICCARDELLA: No, that's not --
MEMBER SHACK: Isn't that what the large gap means, that the nozzle is free to
bend? It's not constrained in the axial?
MR. RICCARDELLA: Yeah, but we don't have the -- this case here -- well, yeah,
a large gap. It's still got some interference.
MEMBER WALLIS: Why does it leap from 20 to .6 in 160 and 180 degrees? Is that
a typo?
MR. RICCARDELLA: No, that's the change in the model from --
MEMBER WALLIS: On the top there, the fourth line.
MR. RICCARDELLA: I understand, yeah. It's the change in the mode from the part
through wall crack to the through wall crack.
MEMBER WALLIS: So by changing the model you make all of the difference in the
world.
MR. RICCARDELLA: Yeah.
CO-CHAIRMAN FORD: Could I make a comment? Again, it's on terms of time management
here. Could I request that the members kind of --
MEMBER WALLIS: We're trying to establish credibility. That's all.
CO-CHAIRMAN FORD: You can take credibility as 100 percent.
(Laughter.)
CO-CHAIRMAN FORD: Okay. Ninety-nine percent. My point is that we requested this
so that the members would understand the approach that was taken, the completeness
of the approach and where we're heading. Obviously this is not finished. There's
no way this is finished, I am assuming, in its entirety.
I just wanted the members to understand the depth of what's being done here.
So, Pete --
MR. RICCARDELLA: We have some very interesting conclusions and observations
on the basis of what we've done, and I'd really like to get to that because
I think there will be a lot of interest --
MEMBER KRESS: Before you go though, one more question.
(Laughter.)
MEMBER KRESS: Your K is a strong function of the residual stresses.
MR. RICCARDELLA: Yes, sir.
MEMBER KRESS: How do you get those?
MR. RICCARDELLA: We have residual stress analyses that were performed, elastic,
plastic residual stress analyses of a nozzle that --
MEMBER KRESS: Based on how it was welded?
MR. RICCARDELLA: Yeah, un-huh, based on weld size. We didn't take into account
that much heat rates and things of that sort. I think standard heat rates were
used, but as I said, I believed there could be an uncertainty as high as a factor
of two on these results.
And they tend not to dominate the probablistic fracture mechanics results because
of the slope of the curve.
Okay. The next is how we use the crack growth data that John Hickling presented,
and here you'll recognize this upper plot with the black as being the fit. This
is the cumulative distribution function for that constant alpha, and the black
points are the heat by heat data.
So each of these points represents the average of those groups of heat. You
remember some heats had 27 specimens; some had one; some had two.
The lower curve with the red data is the integral of all of the data points,
the individual data points, and as you can see that's more conservative. The
mean is a higher alpha for those, and John discussed why that is. It's because
there's more testing that has been performed on the higher rate heats of material
than the other.
What we've chosen to do in our analysis is to use both variabilities, and this
is basically as a result of some comments at the meeting that we had at the
NRC where we look at heat to heat variability, and then we superimpose upon
that within heat variability, and we can specify. For example, you specify 69
nozzles in a head. You could say, well, that head consists of three heats. Twenty
of the nozzles are from one heat, 30 are from another, and ten are from a third
heat.
So we picked from this distribution for the heat, and then we sample again for
the individual nozzles in that heat, and we look at the heat to heat scatter
in that analysis.
And another parameter that we've taken into account, and again as a result of
our interactions with the NRC, is a correlation effect between crack initiation
and crack growth. The comment was that you shouldn't just go in and randomly
for a time to a leakage from the Weibull distribution and then pick a second
completely random parameter because if you have a nozzle that leaks, chances
are that's a bad actor.
So you have higher crack growths in the ones that leaked than the ones that
don't leak. And so what we've done is we've built into our sampling scheme in
the Monte Carlo analysis an ability to correlate the random number for leakage
with the random number for crack growth. Okay?
And this particular slide shows a .8. It's a minus point eight because it turns
out that a high random number for leakage means a long time until leakage, a
high time until leakage. A high number for cracked growth means a high cracked
growth rate.
So if we have a heat of material that's out here in the .8 for crack initiation,
then we're going to sample from this narrower set of data for crack propagation,
and the .8 is an input parameter. We can input zero. They'd be totally independent.
We can input one and be totally correlated. Okay?
So basically it's a knob that we have in our analysis to calibrate or benchmark
our analyses against real behavior. Okay?
MEMBER WALLIS: What are the crosses?
MR. RICCARDELLA: Oh, that's just the randomly generated -- go ahead and put
in the .99. this is the actual spreadsheet that does it.
MEMBER WALLIS: Randomly generated numbers?
MR. RICCARDELLA: Yeah, these are -- let me show you. These are the random numbers
that actually go into the distributions to pick our crack growth rate, to pick
our time to leakage, see.
So if I assume .99 or a very high correlation, basically I'm using the same
random -- it's like using one random number to pick both parameters. They're
very, very highly correlated. Okay?
If I put zero, they're totally independent. So by going from zero to one, I
can span the entire range from no correlation between crack growth, time to
crack growth, and time to initiate in crack growth to very highly correlated
time to initiate a crack growth. Okay?
And then we've got to go back. Thank you.
MS. KING: Sure.
MR. RICCARDELLA: Okay. This shows just some typical results of this analysis.
So this is a typical probablistic fracture mechanics analysis. What I'm creating
is the probability of net section collapse with no inspections for a 602 degree
Fahrenheit head as a function of EFPYs. Okay? Starting in about 20 years, and
I've got several cases here.
First, the difference between these two parameters here represent the difference
between assuming that the head is made up of three heats and the --
CO-CHAIRMAN SIEBER: Could you move closer to the microphone, please?
MR. RICCARDELLA: I'm sorry.
That the head is made up of three heats or 69 heats. In other words, every nozzle
is an individual heat, and that addresses a specific question that came up at
an NRC meeting about is it appropriate to sample each tube individually or should
you be sampling them in groups. It turns out that it really doesn't have a significant
effect.
The other thing we looked at was a log triangular versus a log normal fit of
the data. Did we not go through the -- actually this got a little bit out of
sequence. Let's go to the next two slides.
This is the distribution, again, showing the distribution of that parameter
alpha by heat. The black data points are by heat, and I show a blue curve, which
is the log normal basically that John Hickling presented earlier. The red curve
is the log triangular, is a log triangular fit for that same group of data where
it's truncated at two extremes.
So we don't get into these very, very high crack growth rates in the very tails
of the distribution.
Okay. This is what we're doing for the heat to heat variation, and then the
next slide shows we took the entire population of data and looked at each data
point relative to the mean of its heat, and we developed basically a deviation
from the mean in terms of the multiplication of one for every data point.
And so you see that we get about a plus or minus six multiplier for within heat
variation. So as you go through the Monte Carlo simulation, we say, "Okay.
I have at least 20 tubes in the header out of one heat."
We pick a heat from the previous chart, and for each of those 20 tubes, we sample
from this distribution to say where that -- you know, to get the actual crack
growth rate for that tube, and we correlate that to the time to crack initiation
from the Weibull.
MEMBER KRESS: So for a log triangular, shouldn't you get a discontinuity at
.5 in the slope?
MR. RICCARDELLA: No, I don't think so. At .5?
MEMBER KRESS: That's where the triangle turns around and goes down the other
way.
MR. RICCARDELLA: Yeah, but still, 50 -- there might actually be a -- if you
look real closely there might actually be a discontinuity.
MEMBER KRESS: Okay. Maybe it's just my eyes.
MR. RICCARDELLA: Okay.
MEMBER SHACK: But that's sort of good because those are the sort of two bounding
distributions that you would pick.
MR. RICCARDELLA: Yes, right.
MEMBER SHACK: And you're not seeing all that much.
MR. RICCARDELLA: What we find is about a factor of two, which I think is kind
of within the levels of uncertainty of this type of an analysis really.
MEMBER SHACK: Considering, you know, you'll never determine those tails. So
in one case you've chopped them off and in the other you've let them run to
infinity.
MR. RICCARDELLA: Yeah. Okay. Next slide.
Okay. Now, the real key, I think, to this whole analysis is we've made an attempt
to benchmark the results with respect to the B&M plants, and so what I'm
showing here is -- the previous slides were probably density functions. This
is cumulative probability. Okay?
So this is cumulative probability assuming no inspection for a plant operating
at 602 degrees, like the B&W plants.
This is the cumulative probability of leakage versus time, the cumulative probability
of large circ. crack versus time, and the bottom is the probability of net section
collapse versus time.
And what this slide says is that for that group of the seven B&W plants
operated at approximately this temperature, they had about, at 20 years, those
have about greater than a 90 percent probability of at least one leak, and that's
fairly consistent with the operation. Seven out of seven of them had leaks.
MEMBER SHACK: Of course, since that's how you determine the Weibull, I would
hope it would --
MR. RICCARDELLA: Yeah, right. That's a good observation. But then more significantly,
one out of those seven plants had a large circ. crack. And now when we integrate
the fracture mechanics into it, we're predicting about an 11 or 12 percent probability
of a large circ. crack, and then that drops down to what the actual probability
of net section collapse would have been at that time, assuming no inspections.
Now, as soon as we do inspections, of course, we change that probability of
a large circ. crack.
MEMBER WALLIS: What does net section collapse mean?
MR. RICCARDELLA: Net section collapse basically means nozzle ejection. The same
terminology.
MEMBER WALLIS: The same, okay.
MR. RICCARDELLA: Okay. Now, with that as the methodology, now we've used this
model as a method to basically assess and provide a technical basis for our
proposed inspection plan. Okay?
And the method we've used for this is to, first of all, start with the benchmarked
analysis parameters that I've just described. Okay. So we've somewhat benchmarked
the analysis.
Analyze different plants at various head temperatures, and what we've done is
we've set risk categories based on both the probability of net section collapse
per year and based on the cumulative probability of leakage. Okay?
And then we've also set inspection intervals looking at the effects of inspection
based on probabilities of net section collapse, based on the impact of inspections
on probability of net section collapse.
So I'm going to run through some of the results that we have and then later
this afternoon or this evening, Michael Lashley will talk about the resulting
inspection plan that's resulted from this.
MEMBER WALLIS: Inspections result in a change in the profile though because
you do something as a result of what you find?
MR. RICCARDELLA: Yeah, the assumption is that if you find it you fix it, and
so that particular nozzle no longer has a chance to propagate to ejection. You
fix it or do something to take it out of the mix.
MS. KING: Well, the assumption and the experience to date is that you find it
and you do something about it.
MR. RICCARDELLA: Okay. Just to review the analysis parameters, we've used the
head temperature. We've analyzed ranging from 560 to 605 degrees Fahrenheit.
I mentioned already the Weibull parameters of slope of three with a beta and
a theta of 15 plus or minus six, and it's assumed to be a triangular distribution.
The crack growth rate statistics we've discussed. We're using the log triangular
for both heat to heat and within heat variation.
We've used this cracked growth versus leakage correlation factors. We've used
minus .8 for both the heat to heat and within heat, and --
MEMBER WALLIS: Is this something you thought was real?
MR. RICCARDELLA: Yeah, well, you know, it's kind of the knob that I used to
make it match the results on the B&W plant.
MEMBER WALLIS: I thought it was. There was a dial
MR. RICCARDELLA: It is, in fact, yeah.
Okay, but you know, the real use of this type of analysis is to make apples
to apples comparisons of different things. So I think it's appropriate to pick
a set of numbers.
MEMBER SHACK: It also doesn't seem physically unreasonable.
MR. RICCARDELLA: Yeah.
MS. KING: Right.
MR. RICCARDELLA: You know, if you told me that you wanted me to use log normal
and it doubled the probability, that probably lowered that correlation factor
a little bit because, you know, in the end you want to agree to reality, to
what we've observed in reality, and if reality changes, if we make some inspections
and find some additional unexpected results, we'll have to go back and recalibrate,
I guess.
And then we need some sort of acceptability criteria, and just for purposes
of this inspection plan, what we're using is sort of an acceptable level would
be a probability density function for a nozzle ejection of one times ten to
the minus third, and that's consistent with the most predictions of the consequential
core damage frequency, given a nozzle ejection is also about ten to the minus
third.
So we've got a couple of plots here. This is a plot for a lot of different temperatures,
570 up to 605. The probability of net section collapse versus EFPYs at different
temperatures, and you see two lines on here. You see the 1E to the minus three
that I've talked about as being the acceptability limit. There's also one down
here at 1E to the minus four.
You can see that these tend to jump around a little bit. Let me show you. Go
two ahead to the conversion study.
Here's a convergence study that we did on one particular case with a 600 degree
F. where I've run these with 10,000 -- this is the same thing, probability net
section collapse versus EFPYs, assuming no inspection. I ran them with 10,000
Monte Carlo simulations, 100,000, and then a million Monte Carlo simulations.
Essentially that would take about a ten-hour run to do the million Monte Carlo
simulation.
But you can see that even though you get this jumpy curve, if you pass kind
of a best fit through the jumpy curve, you'd predict about the same time to
one times ten to the minus third. Okay?
All of the cases I showed earlier were run with 100,000, the middle of those
three. Also you see that in terms of the probability of leakage, it has very
little effect, the probability of leakage. Because it's a higher probability
number, basically it converges much faster.
This is the cumulative probability of a leak, assuming no inspections, again,
versus EFPYs for a bunch of different head temperatures. And, again, I've drawn
two horizontal lines on here, one at 75 percent probability of leakage, and
one at 20 percent probability of leakage.
Now, what I've done in the next plot is I've taken the intersections of those
horizontal lines with the results of the analysis and created a locus of basically
a time versus temperature locus of that data. So these upper two curves correspond
to the net section collapse of one times ten to the minus three. That's the
red chain link curve, and the 75 percent probability of leakage in terms of
a time-temperature domain.
The lower two curves represent one times ten to the minus four, probability
of net section collapse, and that just approximately corresponds to a probability
of leakage of about 20 percent.
So what we have here basically is the temperature, the heat temperature, versus
EFPYs of operation. However, as somebody mentioned earlier, some plants have
operated at different head temperatures. They operates for a while at 600, and
then they dropped it to 570.
And so this has been integrated into the number of -- the EFPYs are the effective
EFPYs for those plants that have had multiple head temperatures, assuming that
it has always been operating at the current head temperature.
CO-CHAIRMAN SIEBER: Now, these susceptibilities do not incorporate any knowledge
you might have about the susceptibility of different heats.
MR. RICCARDELLA: No.
CO-CHAIRMAN SIEBER: Okay.
MR. RICCARDELLA: No, we haven't taken that -- this is still generic. Basically
it's a time-temperature, the same type of time-temperature correlation that
Larry was talking about earlier. I've broken it into two, and, in fact, these
are exactly the data points that Larry showed on his plot earlier of the actual
plant.
So we show where the actual plants lie, the 69 plants lie on this time-temperature
domain.
CO-CHAIRMAN SIEBER: I have another question. You have a susceptibility that's
a function of temperature which you've described in everything you've done so
far, but we also know that there is a susceptibility due to the heat. Which
is the more predominant effect as far as determining how long it will be until
section collapse?
For example, all of the leakage we've seen so far came out of one heat, right?
MR. RICCARDELLA: No.
CO-CHAIRMAN SIEBER: No?
MR. RICCARDELLA: A couple of them came out of welds, I guess.
CO-CHAIRMAN SIEBER: All right, okay.
MEMBER SHACK: And there were more than one heat.
MR. RICCARDELLA: There was more than one heat.
MS. KING: There's more than one heat that has leaked.
MR. RICCARDELLA: There is a strong heat-to-heat sensitivity, as there is a strong
temperature effect. Right now I can't say which is more important, but, you
know, the heat-to-heat variability though, that variability is built into the
distributions that we've used in our analysis.
And what we're trying to present here for purposes of the inspection plan is
sort of a summary of the fleet or, you know, a simulation of the entire fleet.
CO-CHAIRMAN SIEBER: I can appreciate that, but when we started out, we used
the temperature data as a basis for ranking the plants and saying these are
the high susceptibility plants; these are moderate; these are low.
And then you put them in order, and that tells the agency who to go after first.
If it doesn't consider the heat data, it's not totally clear to me that we're
capturing everything that needs to be captured to do that ranking.
And this going back to --
MR. RICCARDELLA: But what the problem is is that we really don't have much information
about the susceptibility of the individual heats in the individual plants. So,
I mean, even if -- is that what you were going to say, Larry?
MR. MATHEWS: Well, what I was going to say is by ranking them the way we did,
just based on time and temperature, there's an inherent assumption in that process
that every plant has that same bad material that Oconee 3 had, and it's very
likely that many plants aren't nearly that bad.
CO-CHAIRMAN SIEBER: I guess that's one of my problems, that when you put that
assumption in there, then the ranking is less accurate than it would be if you
took that effect into account.
For example, if you buy a deep draft pump and there was an instance with an
information notice about ten years ago where the heat of some pump couplings
in the shaft was not good, and that became a shut down your plant deal, and
they were able to identify where the bad couplings were depending on when they
were made and who you bought them from.
And so they ought to be able to tell where all of these nozzles came from, right?
MR. RICCARDELLA: Oh, we have information, but we don't have information on the
susceptibility. You know, as we continue to do more inspections and collect
more data, if some form of correlation becomes apparent, we'll take that into
account in the model.
We could adjust this model so that we favor, you know -- so that we could analyze
individual groups of material that are on the bad side or on the moderate side
or on the good side, and if we start to see those --
CO-CHAIRMAN SIEBER: It seems to me that if we continue on this methodology of
inspections and so forth and rankings that you ought to maybe do that.
MS. KING: Yeah, currently we are tracking the inspection data to the heat, but
right now our stance is we don't have enough data to differentiate.
CO-CHAIRMAN SIEBER: To do something real--
MS. KING: And I'm not turning away from that. It's just that at this point we
don't have enough data to differentiate between the heats.
CO-CHAIRMAN SIEBER: Well, if I were in your place rather than mine, I would
be looking to trying to do that, to give me a better picture as to what's going
on as time goes on and you collect the data.
MS. KING: Right.
MEMBER KRESS: Let me see if I can understand the basis behind your one times
ten to the minus three acceptance criteria. If you have that happen, it means
you have a small break LOCA.
MR. RICCARDELLA: Yes.
MEMBER KRESS: And you can't put in one rod.
MS. KING: Essentially.
MEMBER KRESS: Essentially.
MR. RICCARDELLA: Essentially, yeah.
MEMBER KRESS: And that has a conditional core damage probability of probably
ten to the minus three itself.
MR. RICCARDELLA: Yeah.
MEMBER KRESS: So you're talking about one times ten to the minus six core damage
frequency.
MR. RICCARDELLA: Core damage frequency.
MEMBER KRESS: As your acceptance criteria.
CO-CHAIRMAN SIEBER: That's right, or whatever it comes out when you add on all
the other mitigation you get out of the plant.
MEMBER KRESS: Yeah, but that should be at a conditional already.
MR. RICCARDELLA: I guess the other thing, too, is we're really not using that
as acceptance. We're really saying that that's the limit that defines when we
proceed from the moderate risk region into a high risk region, which is using
this to set --
MEMBER KRESS: That's kind of a definition.
MR. RICCARDELLA: -- to set inspection requirements.
MEMBER KRESS: Yeah.
MR. RICCARDELLA: But that's where it comes from, exactly.
MS. KING: And you'll see it hopefully before this evening. We'll get to show
you what these inspection requirements are.
MR. RICCARDELLA: And, you know, we're doing inspections to try to make sure
that we never get to that point. We're starting to do inspections, you know,
even in the low risk regime. We have different inspection levels, but they're
graduated as plants move up from one regime to the other.
So you can see that a high risk model, basically it captured -- there's a total
of nine red points that were leaders. Okay? And all but basically one of those
red points is either on or above our high risk line.
And also I should say that all of these data points are about a year old. So
they're all really going to move up about a year, and actually this data point
here is three plants. There's one right on top of another. So you can't see
them.
The three points where there were inspections that found cracks but no leaks
are the three yellow points, one, two, and three, and then there's a whole group
of plants that have done inspections and found nothing that are shown.
So it has really taken the plot that Larry presented earlier and breaking it
into a two dimensional plot so that you can really see where these plants lie,
time and temperature.
And now the plants progress upward on this line in real time, not in dog years,
but in real years.
(Laughter.)
MEMBER SHACK: What I'm looking at now, when you have those dots, isn't that
a median value for a plant with that temperature?
And if I ranged it up to the 95th percent, as I go through and I vary different
heat assumptions, that low temperature plant is going to get better, and it's
going to get worse, you know.
I've done a bunch of Monte Carlo runs. What's being plotted here? Is that the
median value from that?
MR. RICCARDELLA: This actual data point?
PARTICIPANT: No, the third, the chain link line.
MS. KING: Oh, the chain link line.
MEMBER SHACK: The lines.
MR. RICCARDELLA: The lines are the median results from my Monte Carlo analysis.
The data points are just time and temperature.
MEMBER SHACK: But where would the 95 percentile of the curve be?
MR. RICCARDELLA: I haven't really put confidence bounds yet on the Monte Carlo
analysis. That requires some assumptions about, you know, the confidence in
the various assumptions that occurred.
MEMBER SHACK: I mean, for a given temperature you get a distribution of failures,
right?
MR. RICCARDELLA: Yes.
MEMBER SHACK: Well, you can take the fifth to 95th to that.
MR. RICCARDELLA: This is all of -- this 100 percent of that.
MEMBER SHACK: It's 100 percent?
MR. RICCARDELLA: There's no -- you know, in terms of putting confidence bounds,
I think you'd have to look at uncertainties in your various assumptions that
went into, you know, the analysis.
MEMBER SHACK: This is all of the failures.
MR. RICCARDELLA: This is all of the failures, yeah.
CO-CHAIRMAN FORD: Could I make a suggestion?
MEMBER ROSEN: One question. What would a failure look like? Where would you
plot a low risk plant that had inspected and found the crack? What color would
that be and where would it be? Say it was a 600 degree --
MR. RICCARDELLA: Well, anyone who has inspected and found leakage is a red dot.
Anyone that has inspected and found cracks is the yellow circle.
MEMBER ROSEN: So it doesn't matter whether you're a low risk or a high risk
plant.
MR. RICCARDELLA: No, not in how you plot the individual points. I mean, if --
MEMBER ROSEN: So if I'm to take any comfort from this plot at all in terms of
stuff, you know, if they're falling within the boundaries, because you're by
definition saying if you get a crack or a leak and you're a 600 degree plant,
you're right there.MR. RICCARDELLA: Yeah.
MEMBER ROSEN: And you may need a brand new plant, maybe one of the youngest
plants that --
MR. RICCARDELLA: Well, no, no, no. The probability of leakage for a 600 degree
plant, let's say a 602 degree plant, your probability of leakage hits 20 percent
at about eight years in accordance with this, and then you continue to operate
that plant. It gets higher and higher.
By the time you hit 18 years, it's 75 percent. This is Davis-Besse. It had,
you know, at the time of that inspection about a 75 percent probability of leakage
in accordance with this model.
MEMBER ROSEN: I'm trying to figure out where a point would be on this chart
that would not be consistent with your model.
MR. RICCARDELLA: Oh, if one of these guys comes out as a red triangle, we're
back to the drawing board, okay, or a circ. crack or anything like that. I mean,
then it's reevaluating the whole model.
CO-CHAIRMAN SIEBER: That would tell you your temperature correlation is no good.
MR. RICCARDELLA: Yeah. It might be that the temperature or estimates of that
head is wrong. We don't have absolutely certainty in our estimate of the head
operating temperatures.
CO-CHAIRMAN FORD: Again, Pete, I hate to do this to you, but could you just
move straight to your conclusions?
MS. KING: I guess could we show a couple of slides on the effect of inspections?
CO-CHAIRMAN FORD: Please.
MR. RICCARDELLA: Let's just real quickly show the next one. All we've done here
is to take those same -- that same chart and put on lines of constant EDYs,
which is degradation, and it just shows that's how we get the 18 and the ten
basically, because those are the ones that fall on top of our risk curves.
Okay? All right?
MS. KING: Now we'll go a couple ahead.
MR. RICCARDELLA: Now what we do is I've taken that same analysis, same model
and said here's the probability of net section collapse versus time. This is
run at 600 degrees. So this is actually EDY. It says EFPYs, but in this case
EFPYs equal EDYs.
And at the time that I get to 18 years, which is approximately that one times
ten to the minus three, I assume inspections of various levels. I assume a bare
metal visual, and there's three curves. One is a bare metal visual every refueling
outage. One is every two EDYs. One is every four EDYs. Okay? And what's the
effect of those?
And we made some assumptions about probability of detection, which I think we
should cover.
CO-CHAIRMAN SIEBER: Point, six.
MR. RICCARDELLA: Yeah, we assumed .6, which means --
MS. KING: I think that's hanging up.
MR. RICCARDELLA: I think you've got to go up. No, one more up.
CO-CHAIRMAN SIEBER: We'll trust you it's .6.
MR. RICCARDELLA: Point, six, no, but there's something else I wanted to point
out, was that what we assumed also that was for subsequent exams, if you missed
a leak in a nozzle, we applied a factor of .2 on that. So it's really only .12,
is what we're assuming for subsequent exams of a nozzle.
The comment came from one of our interactions with the NRC that if you do an
inspection and you miss it, it might be because that's a particularly difficult
nozzle to inspect, and you have a higher probability of missing it the next
time.
So we put the ability to input a knockdown factor on the POD or --
CO-CHAIRMAN SIEBER: You can't find it because it's covered up by boron crystals,
right?
MR. RICCARDELLA: Yeah, something like that, or difficult access or tight shrink
fit or all kinds of things. So we think it's a fairly conservative assumption
as to what the POD is, and then we had a second POD set of assumptions for nondestructive
volumetric examination, and that's a curve of probability of detection versus
crack size.
CO-CHAIRMAN SIEBER: Okay.
MR. RICCARDELLA: So if we go to the volumetric, this is the same kind of curve
again. We assume we do the inspection at one times ten to the minus three or
at 18 EDYs, and then what the effects of NDEs at four and eight years are on
that.
CO-CHAIRMAN SIEBER: Do you have one that shows the comparison between a visual
and a volumetric?
MR. RICCARDELLA: The next one sort of shows that.
MS. KING: Kind of, yeah.
CO-CHAIRMAN SIEBER: Okay.
MR. RICCARDELLA: Here's the case of starting the inspections earlier. You know,
we are proposing some inspections for the moderate category. We're not proposing
that people just operate without any inspections until they get to high risk.
We're actually specifying inspections in the low risk and also in the moderate
risk.
These are the moderate risk recommendations, and it's either a visual at two
EDY or an NDE at four. So you can see the effect of the two. Obviously the NDE
is more effective as the large curve.
The NDE at four is more effective than the visual at two.
CO-CHAIRMAN SIEBER: Yeah, that gets back to my earlier point. If you really
want to find them, it ought to be volumetric.
MEMBER ROSEN: It's a question of whether you want to find that or whether you
want them to find us.
CO-CHAIRMAN SIEBER: Well, I think that's well put.
MR. RICCARDELLA: Okay. Let's just go to conclusions.
CO-CHAIRMAN SIEBER: You don't have anything with NDE and visuals at the same
intervals, right? That would be a yes or a no.
MR. RICCARDELLA: No.
MS. KING: No, we do not.
MR. RICCARDELLA: No. We could back it out from the previous two curves if you
want. The NDE is --
CO-CHAIRMAN SIEBER: I think I know enough.
MR. RICCARDELLA: The NDE is much more effective because it's finding -- first
of all, we're using a higher POD for the NDE, and secondly, it's finding things
even before they leak.
CO-CHAIRMAN SIEBER: And it also helps you to some extent to see if you've got
a cavity somehow in the ferritic material that you can't see from the surface,
if you're good enough at looking at it.
CO-CHAIRMAN FORD: I'm going to let the members read the conclusions during their
break time.
CO-CHAIRMAN SIEBER: Our understanding is more important.
MR. RICCARDELLA: This is the key though. I think when Michael gets up later
to present the inspection plan, you're going to see that this is the basic result
of this analysis, is we've got low risk, medium risk, and high risk categories
that correlate to those EFIs, and I've kind of explained where those different
categories come from.
CO-CHAIRMAN FORD: Thank you very much, indeed. I appreciate it.
CO-CHAIRMAN SIEBER: Well done.
CO-CHAIRMAN FORD: We'll recess until quarter past three.
(Whereupon, the foregoing matter went off the record at 3:00 p.m. and went back
on the record at 3:15 p.m.)
CO-CHAIRMAN FORD: Mark, you're up.
MR. KIRK: Okay. Thank you.
Is that working?
MS. WESTON: Yes.
MR. KIRK: Okay. The title of this presentation is NRC assessment of the margin
available at Davis-Besse. My name is Mark Kirk. I'll be making the presentation
for the NRC Office of Research.
What I'll be presenting in the next 40 minutes or so represents the collaboration
of a whole host of people, and I don't think I have all the names on the top
slide.
Wally Norris is another, like myself, is another program manager in the Office
of Research. He manages the work at the Engineering Mechanics Corporation of
Columbus, who has done some of the finite element analysis. I manage the work
at the Oak Ridge National Laboratory under the HSST program.
Of course, Bill Cullen is leading the Davis-Besse effort within the Office of
Research.
Nilerh Chokshi is the head of the Materials and Engineering Branch.
At Oak Ridge, Paul Williams and Richard Bass have been doing the finite element
analysis. At the Engineering Mechanics Corporation of Columbus, the work there
has been led by Gary Wilkowski and Dave Rudland.
CO-CHAIRMAN FORD: Could you just -- we've also got a quorum of people here.
I just want to interrupt for one second just to let everyone know that at the
rate we're going, in case you have to make family arrangements, et cetera, it
might be a quarter to seven or seven o'clock before we're finished, if we keep
up the density of questions.
MEMBER BONACA: Tonight?
CO-CHAIRMAN FORD: Tonight.
(Laughter.)
CO-CHAIRMAN FORD: Sorry, Mark.
MR. KIRK: Okay. And I'll apologize in advance. In order to give you the most
up-to-date information, we've revised these slides since I provided them to
Mike at the end of last week. I do not have handouts right now, but I will.
We'll make them and we'll get them to you. They are probably about twenty percent
changed.
What we'll be talking about today is mainly a discussion of our deterministic
assessment of margins in the condition that existed at Davis-Besse at the time
of the March shutdown.
We'll also be giving you some views on the next steps in this analysis which
include some further refinements of this deterministic assessment and also moving
on to do a probablistic analysis.
The scope of our deterministic assessment was first to asses the margin to rupture
of the exposed cladding left in the condition that existed at the March '02
shutdown.
The next step was to determine how much of either -- how much over-pressure
it would have taken to rupture the cladding in that condition or how much more
wastage would it have taken to rupture the cladding at operating pressure.
Finally, we had planned to assess various weld repair options.
The red text, it's just up here to provide you a perspective of where we are,
well, where we thought we were last Friday. We've had an increased level of
understanding which I think I should say is a reduced level of eposemic (phonetic)
uncertainty regarding our failure criteria. So we are going back and redoing
some calculations, but I think that's all for the better.
We're still working on the middle bullet. The last bullet we are not going to
do because repair isn't being considered at this time.
This slide provides you with an overall perspective of the analytical tools
we've been using. We've been using to different sorts of finite element models.
At Oak Ridge National Laboratory, we've constructed a full 3-D finite element
model where we've got a global model. It includes the specific head geometry
as installed at Davis-Besse. It includes all the control drive penetrations.
That global model, when subject to internal pressure, establishes the boundary
condition on a sub-model which then means that we get a much more refined representation
of the head wastage at least at best we can tell at the time. So this is the
model that we would regard as giving us the answers that are the closest to
reality.
We also have been using an axi-symmetric finite element model. That was constructed
at MC2. Because of the limitations of axi-symmetric modeling, the wastage had
to be modeled as a spherical pit at the top of the head.
Obviously, that's geometrically not a completely accurate representation, but
the reason why we wanted to do that was to enable us to do some quicker parametric
studies about increase growth and so on. Moreover, Gary Wilkowski, who's been
doing the analysis at MC2, has considerable background in modeling of corrosion
damage in gas pipelines and so is familiar with some of the approximations that
is used in that industry.
But in any event, in the end we'll be reporting and relying on the results of
the 3-D model. We've used the axi-symmetric model largely to help guide the
3-D modeling effort and provide quicker results at the time.
This table provides just some details of the analysis and the various inputs
that we've used. The loading in these analyses has been either the design pressure
or in cases where we've tried to calculate the over-pressure margin, obviously,
we've ramped that up.
The temperature has been the operating temperature and we've not considered
any temperature gradients because none exist at operating, at least in any practical
sense.
I'll show you more about the material properties and the local geometry that
we've modeled on the following slide. That's a new slide that is not in your
pack.
On failure criteria, I'm going to give you a little of a now-and-then flavor
because this is the area where we've done some refinements in the last few days.
Up until last Friday, we considered -- or we defined, I should say -- failure
to occur when the average through thickness plastic strain in the exposed cladding
area exceeded 5.5 percent, with the 5.5 percent corresponding to the strain
at the beginning of plastic instability.
That was derived from uni-axial tension data that showed an 11 percent strain
at max. load, and furthered the assumption that failure occurs at the same stress
level under uni-axial and bi-axial loading.
I want to stress that is an assumption that maybe isn't as coupled as well as
it should be to the actual ductile failure mode.
We'd honestly never been completely satisfied with that as a failure criteria
because up until last Friday, we hadn't known of the existence of any better
data to calibrate to. But I'll be discussing how we've changed that shortly.
MEMBER WALLIS: Don't you have stress concentration around the edge where there
is sort of a sharp edge?
MR. KIRK: Yeah. Yes, you do. And that is considered in the geometric finite
element model, yeah.
MEMBER SHACK: Why would you even start with that assumption, Mark?
MR. KIRK: Start with what assumption?
MEMBER SHACK: That failure occurs at the same stress into the uni-axial and
bi-axial loading.
MR. KIRK: If you want the straight and unvarnished answer, because it made the
math work easily. But don't go there too far because everything has changed.
MEMBER SHACK: Okay.
MR. KIRK: Okay. This just shows the material properties that we've been using,
just simply appropriate properties for the RPV steel and for the 308 cladding.
I'm now going to give you a short time history of the geometries we've assumed.
Our first cut at this, when this all hit the fan back in March and the Office
of Research was asked to assist, we had to get some cut on the geometry.
So we took one of the photographs that was taken in the vary initial inspections.
It was a head-down shot of the cavity. At that point, the brown that you see
at the bottom of the cavity, that was water sitting in the cladding.
We used the diameter of the hole as a dimensional reference and simply digitized
the shape of that cavity.
Our current model reflects the results of Figure 13 which is shown in the licensee's
root cause document. It's our understanding that that's the best current representation
of the cavity.
What we've incorporated into our model is a -- I think everybody here has also
seen the companion profile view which shows the nose in the RPV steel. However,
we don't believe or we don't have any reason to believe that that contributes
significantly to the load carrying capacity of the membrane. So we haven't included
that in our model.
Basically, the 3-D model that we're using now has a hole in it down to the cladding
along the green contour.
MEMBER WALLIS: What's the boundary condition on the control rod drive cylinder
there?
I understand the boundary condition around your green line, but what about the
boundary where there's a gray? What's your boundary condition for the cladding
there?
MR. KIRK: I'm not -- I mean it's a -- it's hooked to the rest of the head and
you don't apply a boundary condition there. You apply a boundary condition remote
to the head.
MEMBER KRESS: It's free to move there.
MR. KIRK: It's free to move, yes. It's not constrained. But I'm not sure I'm
answering the question.
MEMBER KRESS: It's hooked at the corners where the green --
MR. KIRK: Yes.
MEMBER WALLIS: So where there isn't green, where that round grey thing is; it's
free there?
MEMBER KRESS: Yeah. It says free-floating membrane, a free-floating area. No
constraint to it.
MEMBER WALLIS: Oh, it can't be. From there?
MR. KIRK: From there, yes. That just expands with pressure.
MEMBER SHACK: No, but the displacement of the cladding is constrained to be
the displacement at the nozzle?
MR. KIRK: Yes.
MEMBER SHACK: It's not free-floating?
MEMBER WALLIS: It's not. It rests on the nozzle.
MEMBER SHACK: And it's attached?
MR. KIRK: This is where the nozzle attaches at. That's correct. Yes.
MEMBER BONACA: I thought the portion of the cladding was exposed within and
beyond the image that you have from a picture taken above.
MR. KIRK: I believe that's what's reflected -- well, these are two different
--
MEMBER BONACA: I understand.
MR. KIRK: Better -- presumably better knowledge going from here to here. I believe
there is -- I should say I believe because nobody is going to band-saw through
this thing and cut it open for all to see.
My understanding from what I have seen -- and I think, you know, yours too --
is that there is exposed cladding back here.
MEMBER BONACA: Exactly.
MR. KIRK: I believe this contour here which I have not outlined is what you
would -- what I'm trying to say --
MEMBER BONACA: Oh, that's what you're seeing?
MR. KIRK: -- is that if you're looking down from the top there's metal here.
This is the position of the nose. That's where the cladding will dispose.
Now what's really there, we still don't know. I think that's fair to point that
out, that in our current calculations -- in anybody's current calculations --
what the actual geometry is is, indeed, unknown. I mean we're getting better
and better representations of it. But I think it is important to point out that
the first order effects that are important is the overall exposed area.
The shape of that, obviously it's different if it is a perfect circle than if
it is along the ellipsoid. Also the details of the thickness, overall thickness
of the cladding and thickness variations.
We don't know all those. You know, those are, to borrow a phrase that I've learned
from our PRA friends, "those are in principle knowable," but we don't
know them right now.
(Laughter.)
MR. KIRK: So we like everyone are proceeding with our best current information
which to my understanding is this right here. But if anybody in the audience
can tell me later about better information, I'd greatly appreciate it because
we're about, as you'll learn later, we're about to embark on finite element
analysis to drive some probablistic calculations. If we can go into that with
a better knowledge of the geometry, that would be certainly desirable.
MEMBER KRESS: I have a little bit of a strange question. Why do you want to
do this?
MR. KIRK: Because my boss asked me to.
(Laughter.)
MR. KIRK: No, you had a serious question.
MEMBER KRESS: Right, seriously. I mean, you're asking -- this is kind of a what-if
question. How close were we to disaster?
MR. KIRK: Exactly.
MEMBER KRESS: Is there some use for that information?
MR. KIRK: In terms of -- do you mean in terms of the probablistic analysis or
the -- we've been doing the deterministic analysis, I think, just to -- my understanding
would be to satisfy that question of how close were we. Were we really close
or were we not so close at all?
MEMBER KRESS: You just want to know that?
MR. KIRK: Yeah.
MEMBER KRESS: Is there some use for that information?
MR. KIRK: Now, going to the probablistic calculation -- and I'll give the short
answer and some of my colleagues in the back can perhaps give a more detailed
answer -- the probablistic calculation is being used as one of the inputs to
NRR's safety determination process.
MEMBER KRESS: The old ASP type thing or?
MR. KIRK: Steve, do you want to take a cut at that so I don't use the wrong
acronyms?
MR. LONG: This is Steve Long.
Significant?
MR. KIRK: Yes. Yes, significant. See, I knew I'd do it wrong.
MEMBER KRESS: Okay. I understand that.
CO-CHAIRMAN SIEBER: The way you can do that is to assume it fails and look at
what mitigating systems were in service and what the failing duct you had, which
comes out to what, three times ten to the minus three or something like that
for CDF?
If you assume the failure frequency is one, that's the first cut.
MR. KIRK: What I'm going to show you is a series of slides that summarize our
current results, and some of these are as current as just this morning. So you
are getting the latest and best.
What the contour plot shows you here is the equivalent plastic straining contours
in the cladding.
We've removed all of the reactor pressure vessel head so you can see what's
going on. We've taken this up to the operating pressure of 2165 psi. At that
pressure we get the highest strain somewhere around about the center of the
wastage cavity, and the peak strain is somewhere between 2.5 and three percent.
We've been going through extensive debates, as I think most of the committee
members are aware, with the industry over what an appropriate failure criteria
is, but I don't think anybody has ever presumed that it would be as low as this.
The finite element model with the best representation of the geometry as showing
us that at operating pressure we wouldn't really expect it to fail. Indeed,
it did not fail.
CO-CHAIRMAN SIEBER: Let me ask you a question. How did you model the cladding
itself? Cladding is not a plate. It is a series of weld stripes, which to me
would seem to be weaker than a solid piece of material that was just a plate
there.
Did you treat the cladding differently--
MR. KIRK: No.
CO-CHAIRMAN SIEBER: -- than you would have as a solid metal?
MR. KIRK: No. Right now -- well, it's weld strip cladding. So we've assumed
-- I mean, it's been modeled as a plate. So you've implicitly assumed that there
are no flaws in it and that you've got no significant lack of inner rod penetration.
CO-CHAIRMAN SIEBER: Do you feel comfortable with that?
MR. KIRK: Yeah. I feel reasonably comfortable with that. The only further modification
that I would think would be appropriate at some point -- and again, this gets
to the question of why are you doing this -- is how refined a model do you want
to get to get a warm, fuzzy feeling that you weren't that close after all.
You might want to include the natural undulations that result from the welding
process. I would personally take the position that I wouldn't want to do that
until I had a lot better picture of what those undulations were. I don't have
that right now.
MEMBER ROSEN: You'd just be making it up.
MR. KIRK: Yeah. Right now I would be forced to make it up. That's right.
MEMBER ROSEN: I'm not sure. You need to be careful about assuming that because
it's weld metal that it's weaker than a plate. There's lots of evidence that
they think it might actually be stronger.
MR. KIRK: Well, just in terms of the --
CO-CHAIRMAN SIEBER: Wait a minute. It seems to me that I've seen weld overlays
on various vessels where it wasn't continuous. I've seen places where the weld
didn't --
PARTICIPANT: Didn't overlap.
CO-CHAIRMAN SIEBER: -- and the undulations actually exist because they are crud-trapped.
That's what makes all these clad vessels, unless they're micro-polished, so
radioactively hot.
MEMBER ROSEN: But would you agree with me that we don't know -- a priori we
don't know whether it's stronger or weaker?
CO-CHAIRMAN SIEBER: I think that you would say it was weaker if you knew exactly
what the weld metal was and the temperature conditions as --
MEMBER ROSEN: But we don't.
CO-CHAIRMAN SIEBER: -- how it was laid down. You would know something about
it, but it would be a guess. It really would.
MEMBER ROSEN: I'm just trying to make the point that we don't know whether it's
stronger or weaker than a model plate because we don't know what the configuration
is (a), and (b) we don't know whether a weld metal deposited that way is, in
fact, weaker or stronger.
MR. POWERS: This is Jim Powers from FENOC.
MEMBER KRESS: The question is: to what detail do you think you all have to go
to with this.
MR. KIRK: One of the things that we will be doing -- and I'll get to this in
a bit in the probablistic analysis -- is we will certainly be including -- because
we know from measurements that were reported in Figure 14 of the licensee's
root cause report; we know there are measurable variations in the cladding thickness.
And so in our probablistic analysis, I can say with a fair degree of certainty
that variations in a uniformed plate model of thickness will be included. Whether
we need to, want to, whether it's warranted to go to the next step and include
the details of the undulations is, indeed, up for questioning.
Like I said, I wouldn't -- I, personally, wouldn't want to do that until I had
a much better picture. By that, I do mean something like a photograph and profilometry
of what's actually there because otherwise I'm just guessing.
CO-CHAIRMAN SIEBER: Okay.
MR. POWERS: This is Jim Powers from FENOC.
We do have some undulations on the surface, but it's relatively smooth. There
is no separation of contact bead to bead. It was a six wire sub-arc application
of the clad. We PT-tested that clad area and found no indications in situ.
So we had some degree of confidence in its continuity.
MEMBER WALLIS: There is a measure of the residual bulging, isn't there, in this?
CO-CHAIRMAN SIEBER: Yes.
MEMBER WALLIS: Does that check your analysis? I mean the actual movement of
the center from --
MR. KIRK: I don't have those figures reported here, but my memory is from an
early analysis that they were -- given the approximations in the analysis and
the difficulty attendant to measuring a set deformation off of initially curved
surface, that if you will forgive the phrase, "they were close enough for
government work."
(Laughter.)
MEMBER WALLIS: That's your predictions, or the measurements?
(Laughter.)
MR. KIRK: Both.
(Laughter.)
MR. KIRK: In this case, the measurements weren't reality either.
I mean, remember those measurements were made in an environment where they were
trying to minimize man REM so it wasn't exactly like somebody got down there
with a micrometer and made a measurement that was good to the mil.
I think we're in the position the piece is now cut out. I apologize because
I don't know where it is. Clearly somebody in this room does. But, you know,
we're in the position of making much less equivocal measurements.
MEMBER WALLIS: Is this going to the Smithsonian or somewhere, is it?
(Laughter.)
MR. KIRK: I don't know.
CO-CHAIRMAN SIEBER: We probably are dwelling on this more than is necessary.
So I think at least I know in my own mind what was done and how it was modeled
and that's good enough for me.
MEMBER KRESS: A strain is a measure in the change in length divided by the original
length.
MR. KIRK: Right.
MEMBER KRESS: Your original length, is that your finite element node that you
use? You get a change in that finite element node?
MR. KIRK: Yes. Yeah.
MEMBER KRESS: Okay.
MR. KIRK: This is the slide there where--
PARTICIPANT: I could put cartoons on there.
MR. KIRK: This is the slide where we've had some significant changes and, I
think, changes for much the better in our predictions of margin on over-pressure.
All the predictions that we've made, that anybody's made obviously depend upon
how you modeled it and what you've assumed for failure.
In particular, the assumed failure criteria, the failure strain makes significant
differences in how much pressure you think you can withstand.
There was considerable discussion given in earlier presentations of this work
that the industry analyses performed by Dr. Riccardella were predicting considerably
higher over-pressure margins than our analyses. It's not hard to see that was
related to differences in the failure strains we were using.
One thing I would point out is that even with our at the time more pessimistic
view of the strain that the material could withstand before failure ensued is
all of our over-pressure margins exceeded the SRV set-point of 110 percent.
So something even with the very pessimistic view that we took initially on what
the material could take, a controlled SRV trip would have happened before we
would have expected the SRV set-point to have been reached, before the membrane
would have blown.
However, as I said, we've been having continuing discussions between ourselves
and the industry regarding the issue of the failure criteria. We've recognized
from the beginning that the failure criteria that we took on was somewhat arbitrary.
Pete pointed out to us -- pointed us back to a paper that he had presented way
back in 1972 at PVP, where experiments had been run on, among other things,
burst discs of 304-stainless steel. I have diagramed the experiment here.
The disc had a thickness of both an eighth inch and a quarter of an inch. It
was a six-inch diameter exposed area, and it was subjected to pressure on the
backside until it ruptured.
Now, to quote Richard Bass, who has looked at this over the weekend, in fact,
if I were to design an experiment to calibrate the failure criteria in my finite
element model, I would have designed this experiment.
So over the weekend, once we finally got the peculiarities of electronic data
transfer perfected and actually got a copy of the paper, Paul Williams and Richard
Bass at Oak Ridge modeled this geometry, which is very conveniently axi-symmetric,
and used it to calibrate a failure model that we would use in the Davis-Besse
analysis.
We believe that these experiments are extremely relevant and appropriate to
this end because the experiments have a similar material, have a similar thickness,
and have a very similar exposed area to the conditions of interest at Davis-Besse.
By calibrating the failure criteria to these experiments, we are able to significantly
reduce our uncertainty in the failure criteria, of course, by referencing the
relevant experiment data.
In doing these analyses, we've reached the same conclusion that was reached
back in 1972, that disc rupture occurs shortly after the finite element solution
fails to converge under pressure loading, of course. What that means, physically,
is that the elements -- we're doing large deformation, large plasticity, finite
element analyses -- the elements have been stretched so far that you can't maintain
-- you can't reach an equilibrium condition.
This, of course, produces what we'll call an NRC failure criteria, which is
much, much closer to that that's been advocated by the industry for quite some
time now.
In exercising this new failure criteria, we're using a new sub-model of the
wastage area just based on our most recent geometric understanding and also
including more refinement through the cladding thickness.
MEMBER KRESS: I have a little problem with that criterion. Doesn't the failure
to converge of your finite element model depend on the size of your finite elements
that you choose?
MR. KIRK: Yeah, yes it does. We've done the studies on that. But the more --
the less refined your model, the stiffer the model becomes. In other words,
our initial model included only one element through the thickness of the cladding.
MEMBER KRESS: Okay. You mean --
MR. KIRK: That --
MEMBER KRESS: -- you only had sort of a surface?
MR. KIRK: Yeah. Yeah.
MEMBER KRESS: That went all the way through?
MR. KIRK: That's right. That model will fail to converge at a lower pressure
than a more refined model. So the -- if you fail to refine adequately, you will
--
MEMBER KRESS: But, how about in the other two dimensions? You can make that
smaller and smaller. On the top.
MR. KIRK: Yeah. Yes, you're right. You're absolutely right that that will depend
upon the level of mesh refinement.
What I'd like to point out is that if you under-refine the mesh, which is the
only error that you can make because it is an inherently discrete model, that
will lead to an under-prediction of the true failure pressure, not an over-prediction.
MEMBER KRESS: You may be right. I am still bothered by having a failure criteria
that's tied to how well my finite element model behaves. It seems a little strange
to me, but I'll buy what you say.
MEMBER SHACK: The system is too stiff. Therefore, I'm going to get less deflection
than I would for a given load.
MR. KIRK: That's right.
MEMBER SHACK: Wouldn't that tell me I'm -- I'm getting less deflection so if
I go -- the strain I'm predicting is really too small, right?
MR. KIRK: I'd have to check, Bill. I think it goes -- oh, what I do remember
is --
MEMBER SHACK: But your going to run the mesh refinement?
MR. KIRK: We're running the mesh refinement. We know that if we have four elements
through the thickness, we get to a higher plastic strain before we can converge
them with one element through the thickness. As of 0800 this morning, we had
a pressure of 3.5 ksi or 60 percent above design without failure, and the model
continues to run.
MEMBER ROSEN: Why wouldn't you let the licensee do any more? They want to get
more and more margins. You know, let them do it. You're done with the problem
as far as I'm concerned.
MR. KIRK: I'll give you a list of people I'd like you to say that too, if you
would.
MEMBER ROSEN: I just said it.
(Laughter.)
MEMBER WALLIS: No, I don't think your done with the problem because the public
is going to ask this question, the newspaper reporters, all kinds of people.
Have you done the ASME diaphragm tests? Have you predicted that too?
MR. KIRK: I'm not familiar.
MEMBER WALLIS: The one you just drew. The one you showed us -- the pictures.
MR. KIRK: Yes. Yes, that's the -- that was the -- I don't have that here, but
those results, we were able to predict the results in the paper.
MEMBER WALLIS: You did a good job on that?
MR. KIRK: Yeah, within ten percent of the -- we systematically under-predicted
the true burst pressure of those experiments by a factor of ten percent.
MR. POWERS: Jim Powers from FENOC again.
From a licensee's perspective, we have a very short presentation that we'd like
to do that shows what we did in terms of optimization of the node, numbers of
nodes for the modeling, as well as a correlation to the disc burst criteria
and shows how we selected our failure criteria. We have about a dozen slides,
if we could respond afterwards.
MR. KIRK: So as I said, these are very new results. This gives you a sense of
the line that we're trying to pursue.
Also, right now the only information that we have on the additional -- how much
bigger the cavity would have had to have been in order to fail comes out of
our axi-symmetric model that was done at MC2.
We haven't yet gotten this into the 3-D model. What we did is we just expanded
the diameter of the pit at the top of the axi-symmetric head until the failure
pressure, and I should emphasize this is a failure pressure at assuming the
old 5.5 percent strain failure criteria. So the newer, better, updated version
should be bigger.
In any event, based on that criteria, we calculated that we needed, in round
terms, two more inches of wastage along the main axis in order to fail at the
operating pressure. Given the changes that I just reflected and our understanding
of an appropriate failure criteria, I would expect that when we do this with
the 3-D model, with the new failure criteria, the amount of the additional wastage
could indeed be considerably more.
MEMBER KRESS: Are you going to convert that into how much time was left before
the --
MR. KIRK: Yes. We do consultation with Bill Cullen. Yes. Yes.
MEMBER ROSEN: Did you get any information about what would happen to the rod
after failure? Would it eject?
MR. KIRK: That's not part of our current analysis. I'll throw that open to anybody
else in the room if anybody -- do you know is anybody considering what would
happen to the rod if this membrane ruptured? Anybody?
I haven't heard about that, but --
MR. POWERS: Jim Powers, from FENOC.
We submitted in our safety analysis of this rod ejection effects. We described
those the last time we came to this subcommittee in terms of the shield above
the rod housing area and lateral loads from jet, cavity loads on adjacent rods,
and the fact that they'd remain in the elastic range and should function properly.
So we had submitted that previously.
MR. KIRK: This is my -- the last slide that I was planning on presenting. It
just gives you a perspective on where we're going. I put on the slide last week
that we were looking at a better definition of the failure criteria. That, based
on the work over the weekend, is now well under way.
On that basis, we intend to recalculate the margin on over-pressure and the
additional cavity growth needed to fail using the new failure criteria and the
3-D model.
As I indicated before, we've begun the FE analysis to support -- to generate
the inputs needed for a probablistic analysis that's needed to support NRs,
and now I've got a wrong again, significance determination process.
MEMBER WALLIS: If your cavity grows enough, then the liner is actually holding
the nozzle.
MR. KIRK: That's correct.
MEMBER BONACA: In fact, a similar question I have is this analysis clearly is
looking at the strain in the material and the ability of what it would take
to rupture.
MR. KIRK: Yeah.
MEMBER BONACA: In reality, during the clean-up of the head, there was work being
done on the nozzle from below and that's when the tube moved.
So I guess the question I have is: how well attached is this nozzle to the cladding,
okay, that would result in that being the weak link?
So, therefore, the cladding probably could have still survived, but the nozzle
would be ejected. I don't understand what caused them to do that.
MR. POWERS: Jim Powers from FENOC.
What caused the rotation is we were going into the repair methodology for the
J-groove weld cracking phenomenon. So we machine-up through that weld and actually
separate from it. Then it wasn't supported up above due to the cavity and it
tipped a bit.
MEMBER BONACA: So the weld was still affected before --
MR. POWERS: That's correct.
MEMBER BONACA: -- they did their work.
MR. POWERS: That's correct.
MEMBER ROSEN: They had already cut it when it moved.
MEMBER BONACA: Okay. I just wanted to know. Okay.
CO-CHAIRMAN FORD: Mark, thank you very much for your -- who's next? Is it the
-- well, how about let's invite Jim Powers?
MR. POWERS: This is Jim Powers.
I'd like Nat Cofie from Structural Integrity Associates to give a short presentation
on what he's done in evaluating this cladding and also comparing it to the disc
burst pressures and give us a quick look at that.
MR. COFIE: My name is Nathaniel Cofie. I work at Structural Integrity Associates.
We've been assisting Davis-Besse in trying to determine the margins, set the
margins in the as-found wastage condition. What I'm here to present this afternoon
is just a very brief summary of the analysis, the failure criteria that we use,
and how we've been able to justify this failure criteria that we're using for
analysis.
We use a three dimensional finite element analysis to build a model.
MEMBER WALLIS: Do we have a --
CO-CHAIRMAN FORD: Yes, it's coming around.
MR. COFIE: We use a 3-D model because this geometry is very, very complicated.
We've tried to explore the possibility of using a 2-D model which would really
make the analysis very, very quick and very ready-available. But the geometry
of the wastage inside that really didn't lend itself to a 2-D axi-symmetric
analysis.
We ran some preliminary 2-D analysis compared with the 3-D results, and they
didn't agree very well. Because of that, we decided to use a three dimensional
finite analysis, which includes the head, the affected nozzle and the adjacent
nozzle, and all the features that would capture the stress and strain distribution
as we subject the head to the pressure loading.
We modeled the entire head and the dummy nozzle and the adjacent nozzles. Of
course, because of the large strains involved in this, we used incremental elastic
plastic analysis.
We used a very conservative stress strength curve in the analysis. And the previous
speaker mentioned 11 percent and 5.5 percent. I'd like to rephrase that a little
bit. It's really not 11.15 percent per C, but the criteria that we used was
basically based on the uniform elongation of the stress-ranked curve. So that
is really the basis for the 11.15.
The criterion that we use in the analysis was that any column of elements in
the finite element analysis, which exceed the uniform elongation, that formed
the basis for our failure criteria. Then we said that we've achieved failure.
We believe this is very conservative because when a column of elements exceed
11.15 percent, or the uniform elongation, there's redistribution of stresses
and strains to the adjacent elements. So using this as a failure criterion to
begin we thought was very, very conservative.
MEMBER SHACK: Suppose you did the more simple minded thing. I mean the uniform
elongation in a tensile test, really it's a maximum load in the tensile test.
So if you do a maximum load in the sphere under pressure and you consider the
thinning of the sphere, you come up with two-thirds of the uniform tensile stress
and the equivalent stress.
Why not that kind of a simple minded argument, where you are basically doing
the same argument, but the thing fails because it is necking faster than it's
work hardening?
MR. COFIE: Yeah, you know you build it -- once you get your large deformation,
you find out using stress as a criterion becomes very difficult. Because a very
small increment in --
MEMBER SHACK: No, it's still a strain criteria.
MR. COFIE: You know, as I will show later, you find that, in fact, when these
evaluations started, the general feeling was that if you run the analysis all
the way to plastic instability, that probably would be very close to the actual
failure. It depends on the --
MEMBER SHACK: That's the equivalent of what you're doing with that kind of an
argument.
MR. COFIE: Exactly.
MEMBER SHACK: Except you're going to do it without making an approximation.
MR. COFIE: Exactly. In fact, later on, I think with the experiment that I --
MEMBER SHACK: But that isn't what you said you did here?
MR. COFIE: Well, I would -- next slide, please.
You find that for the average thickness clad of .297, I've predicted --
CO-CHAIRMAN SIEBER: Could you move closer to the mic, sir? I'm sorry. Could
you move closer to the microphone so the reporter can hear you?
MR. COFIE: Yes. I've predicted that the pressure was 5600, which was greater
than twice the normal operating pressure.
We also ran a case with minimum measured clad thickness of .24. We got 4600
which was also greater.
MEMBER SHACK: But that's for your uniform strain in an element or that's for
your plastic instability?
MR. COFIE: No, that is for the uniform strain, elongation.
MEMBER SHACK: I see.
MR. COFIE: So even with this conservative failure criteria, we got failure pressures
which were twice, greater than twice the normal operating pressure. Of course,
if we had used the instability load as a criterion, that would have been even
greater. I'll show you those.
But this criterion came under a little bit of question because it was slightly
based on engineering judgement, engineering judgment, but I believe that based
on the fact that once you reached the uniform elongation, there's free distribution
of stresses and strain. I believe strongly that that was very conservative.
But now that -- to prove, that fortunately we got this burst test that were
run by PVRC somewhere in the early '70s. And Pete took that results. Peter Riccardella
took those results and did an analysis of those.
So we had the test results available to us. So we used that basically to test
the failure criteria that we've used to see how reasonable it is and whether
it is conservative enough application to this wastage problem that we dealt
with.
Next slide, please.
This didn't come out very well, but this is basically the three dimensional
finite element model that we use. It takes a very sophisticated finite element
model.
We have the -- this is the wastage area right here. That doesn't show very well
on this slide. This is the nozzle associated with the wastage.
We modeled the four adjacent nozzles so that we catch the ligament effect. Initially
we thought that we could get by without making this thing too complicated, but
we wanted to get you all the details. You know, once you have some adjacent
holes in the neighborhood of this area, we thought that could affect them. So
we modeled adjacent models to catch the ligament effect.
This model has a total of about 6,000 elements. Through the cladding, we had
six through wall elements. Because of that, the became a very humongous model
which took days to run. We wanted to do it right, to make sure that we get results
that we can rely upon, are very, very reliable.
Next slide, please.
This is a summary of the analysis results. The original footprint, which is
20.5 square inch with an average thickness clad, using the criterion based upon
the uniform elongation, the predicted failure pressure was 5600.
For this case, we also went as far as to instability, and the instability pressure
was greater than 8000. If I remember, it was 8,125. It was greater than 8000.
We also looked at a case with the minimum measured thickness of .24 and the
original footprint. The predicted pressure based on the uniform elongation criteria
was 4600.
For this case, we really didn't go to instability because this was failure criteria.
Therefore, we just run a little bit greater than this pressure. We know that
the instability pressure is 48 -- greater than 4800.
If I were a betting man and you asked me what would be the instability pressure,
I would say probably it goes up to about 7,000.
We also did another analysis to look at what is the failure pressure if, indeed,
we have a larger footprint, twice the area that was associated with the wastage.
And in this case we've got a predicted failure pressure greater than 2750.
Once again, we stopped this just around about 3,000 because we had used a 2-D
model to basically benchmark against a 3-D model, to predict when we'd get to
about 2750. So we didn't run this under any pressure greater than 3,000 or so.
I believe that the instability pressure for this one is also greater than 4,00
psi.
Next slide, please.
MEMBER BONACA: That's the question I had. Can you comment on the stuff used
5.5 percent strained?
MR. COFIE: Right.
MEMBER BONACA: Okay. It seems to be a key to the difference as one may see.
Could you comment why you use 11 percent?
MR. COFIE: Like I said, 11.15 percent was basically the uniform elongation.
The idea of using that as a failure criteria, that once you reach 11.5 percent,
once you reach the uniform elongation, you start to get necking (phonetic).
That is the first onset of instability, but that's not necessarily the failure
pressure.
I understand the 5.5 percent was used by the staff's contractor because they
were trying to account for the so-called tri-axiality factor associated with
the fact that this is sort of uni-axial loading, but it's a bi-axial, tri-axial
state loading.
Unfortunately, you don't apply the tri-axiality factor to the uniform elongation.
I believe, probably, if that has been explored a little bit, they would have
probably done something a little different. I think that is the difference between
the two failure criteria that we used.
I brought this slide only to show that the uniform elongation associated with
the stress-strain curve that we used is very conservative. There's a whole lot
of reference in the literature, a lot of NUREGs and EPRI reports that have reported
various elastic -- various stress-strain parameters for weld metal as well as
base metal, stainless steel.
Here is the uniform elongation for the base metal. It is pretty large. Our range
is about 45 percent. So make-up weldments, SAWs, the average for all the data
that would have to got in this reference is about 25.7, 25.7 percent. Most of
the data is greater than 20 percent with only two of them less than 20 percent.
This is the data for SMAW weldments, all of them greater than 20 percent, with
the average of about 30.7 percent.
If you combine both populations, both weld populations, both SAW and SMAW, the
average elongation is about 27.3 percent.
MEMBER SHACK: But, I thought Davis-Besse when they were last here told us those
came from measurements by Framatome, the 11 percent.
MR. COFIE: No, that's not exactly correct. The 11 -- the stress-strain curve
that was used was basically obtained in the lurch and one of these handbooks.
At that time, that was conservative enough that we chose to use that one as
the basis for these evaluations.
But, no, there was no measurement made on the Davis-Besse --
MEMBER SHACK: No, not on the Davis-Besse. A test on weld metal, on cladding?
MR. COFIE: Yes, it was based on the test on that, but just obtained from the
literature. Okay?
MR. POWERS: This is Jim Powers from FENOC.
Steve Fyfitch was there at that meeting and indicated that it came from Oak
Ridge data in the handbook for the specific 308 material stress-strain curve.
So it was the best data we had specific to the weld material.
MR. COFIE: So, in any case, you look at all this data and compare it to the
data that was used for the evaluation, and you find that we are really on the
very conservative side of what is reported in the literature.
Next slide, please.
Okay. From this slide going -- I'm going to just describe, basically, the disc
burst test that PVRC -- disk burst test which was performed in the early '70s
which was basically used to determine the reasonableness and conservative --
the conservatisms involved in the failure criteria that we used.
CO-CHAIRMAN FORD: This is just essentially the test that Mark just --
MR. COFIE: Yeah. This is just the test that Mark had presented.
CO-CHAIRMAN FORD: Could you just highlight any differences in your approach
versus that which he did? Just for sake of time just highlight any differences
you may have in your approach and --
MR. COFIE: Just to save time, this is basically, you know, the three geometries
that I we did.
Next slide, please.
These are basically the properties associated with the materials of the disc.
Next slide.
Well, this slide also did not show up very well. What we did was that we looked
at various through wall elements, four, eight, 12, just to look at the effect
of through wall elements on the analysis results.
Next slide, please.
Okay. We also -- the slide you just saw was the axi-symmetric model. This one
is a three dimensional model very, very similar to the model that we used for
Davis-Besse. We used the same element types so that we get one-to-one comparison.
Next slide, please.
Now, these results show the effect of through wall elements versus the predicted
pressure or the predicted failure pressure. As you can see from six elements
to -- six elements onwards, there is essentially no difference in the predicted
failure pressure.
So any time you use through wall elements of six and above, basically, you get
essentially the same results.
In doing Davis-Besse's model, we used six through wall elements to decline.
Next slide.
MEMBER WALLIS: Why do you have two curves here? Why are those two different?
MR. COFIE: Well, there are two different because if you remember there's --
I presented two models. One is a 3-D model and one is an axi-symmetric.
MEMBER WALLIS: It seems to me that all the points are for the axi-geometry or
am I misreading the figure?
MR. COFIE: Well, there are -- in the PVRC test there were three different geometries.
MEMBER WALLIS: But why are some at 7,000 and some at 14,000?
MR. COFIE: There are three different geometries.
PARTICIPANT: Some of them were one inch thick and some of them were a quarter
of an inch thick.
MEMBER WALLIS: Okay. Okay. Then that's why it's twice the pressure.
MR. COFIE: Exactly.
MEMBER WALLIS: Okay. I understand.
MR. COFIE: Next slide, please.
Well, here is the typical stress-strain associated with the disc analysis that
we did. Remind you everything is essentially membrane. You know, at very high
pressure this is almost like a balloon, very, very high membrane stress. Right
at the edge here, we have some venting stresses here.
Next slide.
Okay. This is a summary of one of the analysis results. This is the total equivalence
strain of -- when makes it strain, this is pressure for one of the analysis
that we did for the disc burst test. We flooded both the top level and the bottom
strains as a function of pressure. This is how the outage behaves.
Okay. This really is the point where right at the end of the evaluation or the
end of the pressure increment is where we reached instability.
So the instability pressure associated with this particular test was about 14,000
psi compared to a test pressure, a test burst pressure, of 15,000. So even at
instability, we've predicted that we are slightly below the burst pressure obtained
in the test.
Now, based on the elongation, based on the uniform elongation criterion that
we use for Davis-Besse, this is where we would have predicted failure. We would
have predicted failure right around about 11,000 psi, which is, of course, significantly
below the test burst pressure.
MEMBER SHACK: Maybe I'm reading something wrong here. As I read from the paper,
it says all the center line failures occurred at approximately the same strain
level, 35 percent.
MR. COFIE: Well, don't forget that when Pete did this analysis, when this analysis
was done, it was done with only one through wall element. Really this analysis
is a refinement of what was done in 1972.
MEMBER SHACK: Oh, so the 35 percent is not a measurement?
MR. COFIE: No.
MEMBER SHACK: It's an analysis?
MR. COFIE: It's an analysis.
MR. RICCARDELLA: Yeah, this is Pete Riccardella from Structural Integrity.
Yeah, you have to recognize that they did that analysis with the tools that
existed back in 1972. So you really have to ignore some of the analytical predictions
there. We've updated that analysis with today's tools. So that 35 percent represents
sort of an old estimate.
MR. COFIE: Right. If you read the paper, I find that one through wall element
was used. This had about twelve through wall elements. So this is a much more
actual analysis that we've done.
So this tells you that the criteria now we're using is very conservative compared
to the test results.
Not only that. The instability pressure also predicted pressures which are significantly
-- well, not significantly, but slightly below the test pressure.
So really one can argue that you could go to instability and that would be a
very, very good criterion to use to predict the best pressure.
Next slide, please.
MEMBER SHACK: Unless it falls at an edge, right?
MR. COFIE: Well, even that fail at the edge, you know, also predicted the same
thing.
Here's a summary of all the analysis that we did on the burst test. Here is
the burst test results.
Where is instability? We find that instability is very, very close to the burst
test results. This is -- this are the results based upon a uniform elongation,
and you can see that is conservative compared to the burst test.
So of all the analysis that we did to find out in all cases, the criteria that
we've used for Davis-Besse is very conservative.
This simple analysis that we've done has proved beyond any reasonable doubt
because now we have got some work as data that the criteria that we've used
is conservative.
So, anyway, in confusion, I would say that what we've done for Davis-Besse,
you know, we've done a very conservative analysis. We've used very good finite
element models, 3-D finite element model. Like Mark said, we've used a lot of
through wall elements to the cladding. We've also, basically, tested the criteria
against known burst test results to show that it is very conservative.
CO-CHAIRMAN FORD: Thank you very much indeed. We appreciate that.
MR. COFIE: Thank you.
CO-CHAIRMAN FORD: As I understand it, now we've got three presentations, one
by FENOC and then one by you, Jim. And then one by you, Larry. They are all
scheduled for one hour each.
If I could ask you to please look at your presentations and try to make them
three quarters of an hour each, I'd appreciate that very much.
(Pause in proceedings.)
MR. POWERS: Okay. good afternoon. I'm Jim Powers. I'm the Engineering Director
for First Energy at the Davis-Besse plant.
This afternoon, we're going to do a brief update to the ACRS ON where we stand
with the situation at Davis-Besse. I brought along with me a number of individuals.
You will recognize Mark McLaughlin as our Field Project Manager for work on
the head.
Bob Schrauder is our Director of Life Cycle Management and is responsible for
the replacement head project that's ongoing.
And Steve Loehlein is our root cause lead investigator. He'll give you an update
on what's transpired in the root cause area.
CO-CHAIRMAN FORD: Thank you.
MR. POWERS: Okay. So with that, let me turn it right over to Mark, and he will
give us a description of field activities.
MR. McLAUGHLIN: Okay. Good afternoon. I will definitely try to be brief.
The first slide -- the next, keep going -- okay. The one thing that I wanted
to point out, you guys had seen this slide before. I just wanted to point out
the access that we had to do our inspection, and this kind of leads into the
root cause report that will be coming up.
These are what are commonly called mouse holes and those were five by seven
and they were installed in this lower portion of the service structure.
Next slide, please.
You've seen this nozzle depiction many times. The only thing that I wanted to
point out is that on a Babcock & Wilcox reactor head, this is a gasketed
joint with no seal weld. When these leak, the path that the borated water takes
to get down to the head would be twofold.
One, it could drip down onto the insulation, and there is an eighth of an inch
gap between the nozzle outside diameter and the insulation, or a nozzle, an
adjacent nozzle in this area could spray onto this, and we have observed both
of those types of leakage, And then it flows down and through the gap.
Next slide, please.
I wanted to update you with two things on this slide. Nozzle number two, we
originally reported that there are eight axial flaws. There are actually nine
axial flaws with this. That also brings the total number to six through wall.
The other thing, if you notice nozzle number 46 we say has no flaw indication.
However, there was a shadow. What we've done since we were here last time is
we've cut the nozzle up into the shadow region. We did a visual inspection,
as well as a dye penetrant inspection. I guess the results are that we really
don't see any reason why that shadow is there. There is no leak path present
and there is no significant corrosion.
Let's skip this next one.
I guess the big thing that we've done since we were here last is we did perform
the abrasive water jet cutting of the cavity. The cavity has been removed. What
your seeing here is the water jet tool. This is a mock-up. We mocked-up -- performed
two mock-up cuts prior to performing this cut on the head.
Next slide.
This is the actual cutout on the reactor pressure vessel head at Davis-Besse.
You notice nozzle number 11 would have been in this location. We used nozzle
number 11 as the entrance point so that we wouldn't do any damage to the weld
material around nozzle number three to preserve it for experiments.
Next slide, please.
This is another view using a remote camera underneath the head of the cutout.
Next slide.
This is an actual view of the cavity that's been removed. It shows the lithium
fixture and the as-removed was about a 17-inch diameter.
I just wanted to update you on the sample plans of what we have. Phase one was
various boron samples that we had collected from the -- on top of our head.
We do have a draft report from our contractor who's been analyzing those.
We've found what we expected. There's significant boron, iron, and lithium.
There's also some traces of nickel and chromium which is probably from either
the nozzle material or the weld material.
Phase two is currently in analysis. That is boron and material samples from
the removal of nozzle number two. So that may give us some boron samples in
the actual annulus region.
Phase three, we are currently working with the staff to determine what type
of testing and experiments we want to do on the actual nozzle number three,
the actual nozzle from number two, and the cavity.
Somebody was asking about dimensions earlier. All of these samples are down
in Lynchburg, Virginia. We are arranging a trip down there within the next two
weeks. Anyone who would like to go see the cavity, touch it, and measure it
as much as they want, it is available.
PARTICIPANT: Keep the ALARA advised.
MR. POWERS: It is a much lower dose.
MR. McLAUGHLIN: Yeah, the dose is significantly lower now.
MR. POWERS: Yeah.
MR. McLAUGHLIN: This is a picture of looking in the cavity after it was removed.
You can see in the under-hung portion, and I think you get an excellent view
of the cavity itself as well as the exposed cladding.
The cladding looks brown because it still had the abrasive on it from the abrasive
water jet cutting process.
Next slide.
What I wanted to show you here is -- the last time that we had talked to you
there was some discussion about a detachment or corrosion between the cladding
and the base metal around nozzle 11. What this is, this is the J-groove weld
for nozzle 11 and you can see the opening where we entered to do the cut through
nozzle 11.
I performed an inspection. The surface is too rough to do a dye penetrant test
at this time. However, there is no evidence of cladding detachment or a corrosion
in that region between the cladding and the base material.
That's all I have. Are there any questions as far as updates from the field
activities?
(No response.)
MR. McLAUGHLIN: Hearing none, I'd like to turn it over to Bob Schrauder, who
is going to discuss the replacement of the reactor pressure vessel head at Davis-Besse.
MR. SCHRAUDER: Good afternoon. While Jim and Mark were busy attempting to repair
the reactor vessel head, I started out early on in the process looking for a
potential replacement head for the vessel.
We looked at several options, one of which was to look at -- we do have a new
head ordered for Davis-Besse that was scheduled to arrive at our plant during
the first quarter of 2004. We looked at accelerating that schedule.
We also looked at potentially purchasing someone else's place in line, if you
will, that had another head already ordered that would be compatible with the
Davis-Besse vessel also. Those, the earliest one coming out there that we could
find that was compatible was in the third or fourth quarter of 2003.
Both of those being manufactured, ours and the next one in the pipeline, if
you will, were already on an accelerated schedule. So we were not going to be
able to do much with the schedule of getting a new head in here much before
2004.
So I then began to look at what was already available in the industry. We found
two heads that were compatible with Davis-Besse. Rancho Secho had a plant that
had operated for a while and, as you know, has been shut down. And then at Midland,
one of the two heads in that unit was still on site there.
We looked at those two options. We quickly zeroed in on the Midland head.
Next slide, please.
The Midland head -- both heads like I said would fit with some very minor adjustments.
We thought Midland was the clear choice for two reasons. One, it was a lot closer
to us. It's in the neighboring state right in Midland. We have to just bring
it across the state line and bring it down to Davis-Besse.
It is readily available from the perspective of it's sitting in a commercial
setting, if you will. It would be a commercial kind of construction job to go
get it versus the Rancho Secho head which is, although not an operating nuclear
plant, it is still a nuclear plant. That head, because it was used, was contaminated
which complicated both any modifications we might need to make with it and significantly
complicated the transportation needs for that.
CO-CHAIRMAN SIEBER: Well, the Rancho Secho head was still installed, right?
MR. SCHRAUDER: That's correct.
CO-CHAIRMAN SIEBER: So you would have had to cut a hole in their containment
to get it out?
MR. SCHRAUDER: Well, actually the Rancho Secho head will fit through their equipment
hatch.
CO-CHAIRMAN SIEBER: Oh, yeah?
MR. SCHRAUDER: The Midland head would not.
CO-CHAIRMAN SIEBER: Will it fit through yours?
MR. SCHRAUDER: No, it will not.
CO-CHAIRMAN SIEBER: All right. You can tell us about that later on.
MR. SCHRAUDER: Yes, that is in the presentation. We'll get to that.
PARTICIPANT: They're not going to fold the head.
CO-CHAIRMAN SIEBER: Cut the head in half and put it in the containment.
MR. SCHRAUDER: The other head at Midland is, by the way, cut in half. So that
one was not usable.
(Laughter.)
MR. SCHRAUDER: This slide shows some of the similarities between the Midland
head and the Davis-Besse head. They were both fabricated by Babcock & Wilcox
in the same period of time to the same ASME boiler pressure code edition and
addenda.
Now we have the records for the Midland head. We know that during construction
that head was accepted for use by Consumers Power. It was signed off by an authorized
nuclear inspector and identified as an acceptable ASME component.
It was, in fact, as all of the B&W plants were, it was hydroed before it
was shipped to the site. It shows the hydro was there at 31.5 pounds. As you
know, Consumers canceled the original plant back in the mid-1980s. Since that
time, that head has been on the head stand inside the containment.
CO-CHAIRMAN SIEBER: You knew what heat and nozzles came from?
MR. SCHRAUDER: Yes, sir. That's right around in the presentation and we'll get
to that.
CO-CHAIRMAN SIEBER: Thank you.
Did you file a Part 21 related to the nozzles that were susceptible in your
plant?
MR. POWERS: No, I don't think we've filed a Part 21 as of yet. But we've had
discussions on that issue.
CO-CHAIRMAN SIEBER: I think that would be a good discussion to have amongst
yourselves.
MR. SCHRAUDER: Because of their technical expertise and because of the fact
that they had access to all of the records on this heat, we hired or we brought
in with us a partner Framatome. Framatome actually purchased the head for us
from Consumers. They purchased it as a basic component.
They'll verify its usability. They'll compile for us the code data package which
they have the records for. They'll disposition any non-conformances on that
head and then will sell it to us as a basic component for use at Davis-Besse.
CO-CHAIRMAN SIEBER: Is that an assembly or is it just the head? In other words
are the control rod drive mechanisms already installed?
MR. SCHRAUDER: The control rod drive mechanisms have been removed and somebody
else owns those.
CO-CHAIRMAN SIEBER: Okay. You're going to use your old ones?
MR. SCHRAUDER: That's correct.
CO-CHAIRMAN SIEBER: Thank you.
MR. SCHRAUDER: In the process of this, Framatome will also reconcile the design
requirements of the Midland plant to the Davis-Besse plant. Those design requirements,
again, are covered over the next couple of slides.
Of course, Framatome will do these activities under their quality assurance
program, including responsibility for Part 21 reporting.
The next slide is simply a pictorial that you can relate to as we talk about
some of the similarities and differences on this head.
This next slide shows that this head and the design is essentially identical
to Davis-Besse. They were both 177 plants. The materials of construction you
see there are virtually identical. Even the closure head flange there is really
the same material, the same specs. for that material.
The design pressure and temperatures for both reactors was identical, 2500 pounds
of pressure and 650 degrees.
MEMBER ROSEN: What does the dash 64 mean on the closure head flange?
MR. SCHRAUDER: Actually that's an A50864, and that's an ASTM material. They're
the same material essentially. One is an ASTM code.
MR. POWERS: Go ahead, Steve.
MR. FYFITCH: To answer his question -- this is Steve Fyfitch, Framatome -- the
64 is the date, the year. So it's the 1964 edition of the ASTM code or the ASTM
specs.. Excuse me.
MEMBER ROSEN: But the materials are the same is the answer I got.
MR. FYFITCH: That's correct.
MR. SCHRAUDER: The next slide that answers the question of do we know the heat
materials on this head. In fact, we do. Sixty-eight of them are from the specified
heat there, M7929. And one is from M6623.
What happened was in the manufacturing, the putting together of this head, there
was one nozzle 7929 that had had a problem, and the other nozzle came from the
canceled Bellafont (phonetic) unit. So that's why there is one nozzle that's
the same.
Neither of those two heats of materials has any industry experience. We do know,
though, that they look to fall right in the middle of the pack by way of yield
strength for those heats. But there is no industry experience on them.
The control rod configuration and the alignment is the same on that head as
it would be for Davis-Besse. So, geometrically, it's very nearly the same or
physically, its characteristics are the same.
There are a few minor physical things that we have to do to the head. The picture
that is shown here is the key-way. The key-way fits into the reactor vessel
itself and it makes sure that the head is precisely aligned to the vessel for
latching your control rods and your control rod interface.
There's two surfaces. You see that one and then the other one would be on the
inside there. There's two surfaces for each of the four key-ways that you have
to be concerned about getting your fit. Four of the eight surfaces needed to
have some slight machining to precisely fit on our reactor vessel head; to the
tune of about five mils., we had to machine on those.
Also the control rod drive flange itself that is on the nozzle that the control
rod drive mechanism flanges to has an indexing pin on that, too. There's two
locations that you can have your -- that that's indexed too. As you might figure,
they used the opposite hole that Davis-Besse does.
So we have to take the plug out of their indexing pin in that location, put
it in the other location, and then we'll have that set-up to index for our control
rod drive mechanisms.
The next slide shows another physical difference on this head and ours. The
Davis-Besse head has the O-ring which is the sealing ring for the head to vessel;
is 0.5 inches on Davis-Besse. On the Midland one, it was 0.455.
CO-CHAIRMAN SIEBER: You have two O-rings?
MR. SCHRAUDER: Yes, sir.
So, those two O-rings -- we have done the analysis to show that it will effectively
seal in the groove that we have on our vessel. Of course, we will be able to
demonstrate that with the leak-off capability on that head. We will be able
to tell if there's any leakage between those seals.
CO-CHAIRMAN SIEBER: But the clearance between the vessel flange and the head
flange will be slightly smaller, correct?
MR. SCHRAUDER: Well, the crush is fine on it.
CO-CHAIRMAN SIEBER: Okay.
MR. SCHRAUDER: So the sealing surface that you need, both of those surfaces
we show will have full contact and it won't be an issue.
So we're manufacturing the new O-rings to 0.455 that will fit inside the groove
for the Midland head.
The next slide, again, is a pictorial that you can refer to for the next series
of slides that I'll go over, which describe the nondestructive exams that we'll
do on this head to verify that its stay in Midland, since they canceled that
plant, has not had any deleterious effects to it.
We did three types of exams on this head or will do three types of exams. One
is to supplement the ASME code data package. One is our pre-service IS exams,
and then we did some additional nondestructive exams to verify that, again,
there was no deleterious effects to the head from the period of time that it's
been sitting in Midland.
This first page shows the examinations to supplement the ASME code data package.
I should mention that with supplement, although we had a signed off code data
form, we did not have the film of the radiographs for this head. We had indication
and sign-off that they had a successful radiograph both on the dome to flange
weld. This is a two-piece forging for this head. We didn't have that radiograph
film and we didn't have the radiograph film of the nozzle, the flange to nozzle.
So we're re-radiographing both of those. In fact --
CO-CHAIRMAN SIEBER: You need to have that.
MR. SCHRAUDER: We have completed the radiograph on the large dome to flange
weld. That radiograph did prove to be very sound.
We'll do a series of visual exams, just to verify there is no obvious problems
on the seating surfaces and the grooves in this head.
And I discussed the radiographs that we'll do. And we'll also do a PT examination
on the J-groove welds.
CO-CHAIRMAN SIEBER: And a visual on the inside cladding to make sure it's all
there?
MR. SCHRAUDER: We're going to do some liquid penetrants on the surface of the
and the repaired areas of the clad, of underneath.
CO-CHAIRMAN SIEBER: Okay.
MR. McLAUGHLIN: The cladding is all there.
CO-CHAIRMAN SIEBER: Okay.
MR. McLAUGHLIN: Yes.
(Laughter.)
CO-CHAIRMAN SIEBER: Well, sometimes it isn't, you know.
MR. McLAUGHLIN: Oh, really?
CO-CHAIRMAN SIEBER: Yes. Yes, sir.
MR. McLAUGHLIN: I was up Friday and inspected it. The cladding is all there
in this head.
CO-CHAIRMAN SIEBER: PT is pretty hard to do on a welded surface that's not cleaned
up, right? Dye penetrant?
MR. SCHRAUDER: The --
MR. McLAUGHLIN: It won't be a problem on this head. When I was in there, I'm
not sure what process they used. They must do some grinding on it because the
inside diameter of the head is very smooth.
CO-CHAIRMAN SIEBER: Oh, it is?
PARTICIPANT: The cladding?
MR. McLAUGHLIN: Yes, the cladding is smooth, as well as it was on the Davis-Besse
head.
CO-CHAIRMAN SIEBER: Oh, okay. That's also not always the case.
MR. SCHRAUDER: The next slide shows the pre-service inspections that we'll do:
magnetic particle exam with the flange to dome weld, an ultrasonic on that same
weld, and an electric penetrant PT exam of the -- this has the peripheral CRDM
nozzle to flange welds, the ones on the peripheral. That's what's required by
code. Our intent is to do all of them that we can get to. We believe that we
will be successful in getting to all of them. We will certainly, at least, meet
the code requirements for that, and our expectation is to do PT on all of those.
The next page just shows the additional nondestructive exams we'll do, chemical
smears to assure that it meets the proper class cleanliness. A baseline UT we
will do on all of these nozzles so that if we do UTs in the future, we will
have something to compare to. We'll know whether there was any indications in
these nozzles early on.
CO-CHAIRMAN SIEBER: What kind of packaging was the head stored in?
MR. SCHRAUDER: It was not stored in any packaging.
CO-CHAIRMAN SIEBER: It was not covered?
MR. SCHRAUDER: No, it's not.
CO-CHAIRMAN SIEBER: It's in a building; is that correct?
MR. SCHRAUDER: The CRD nozzles did have some covering on them, but that was
about all that was covered. It's in the containment building. That's correct.
CO-CHAIRMAN SIEBER: So you have to cut a whole in that one, too?
MR. SCHRAUDER: That's correct.
MR. McLAUGHLIN: That's correct.
MR. SCHRAUDER: And that's coming up next on how we're going to go about getting
this.
I should mention -- I meant to mention this earlier -- our intent is to use
this head, put it on now. We'll use it until such time as we replace our steam
generators, which is currently expected to be 2010 or 2012, in that time frame.
So we are maintaining our place in line with our new head. We will get a new
head and we will replace it again when we open the containment up again to replace
steam generators.
CO-CHAIRMAN SIEBER: Now, why would you do that? For material change?
MR. SCHRAUDER: That's right. This head obviously has the same material on it,
the same susceptibility.
CO-CHAIRMAN SIEBER: You moved the 690?
MR. McLAUGHLIN: Correct.
CO-CHAIRMAN SIEBER: Do you folks know a lot about 690 as far as the nozzle database?
MR. McLAUGHLIN: From what I understand, I don't believe there is a large nozzle
database.
CO-CHAIRMAN SIEBER: Or any database, right?
MR. McLAUGHLIN: Well, there would be some in France.
CO-CHAIRMAN SIEBER: Okay.
MR. McLAUGHLIN: But they'd be young.
MEMBER ROSEN: What you want to do is take the head you take off, this one, and
put it someplace and protect it.
CO-CHAIRMAN SIEBER: Well, I don't know. You have a whole --
MEMBER ROSEN: Just swap back and forth.
(Laughter.)
PARTICIPANT: Well, just don't take it out to the dump.
MR. SCHRAUDER: Well, as a matter of fact, our intention is to dispose of it
shortly after we take it out of containment, if it is categorized as a Class
A alpha waste.
MR. FYFITCH: Let me just add a point. John, I don't know where you are going
with that question. This is Steve Fyfitch, again from Framatome.
The 690 has been in use now in steam generators for a number of years, and on
France for the nozzles on the head, they've been replacing heads since the early
'90s. So now they are almost nine or ten years in service.
By the time Davis-Besse replaces theirs in 2012 or 2010, it will be almost 20
years. So there will be a large database of experience by that point.
CO-CHAIRMAN SIEBER: Well, my only comment, I guess, is I started in this business
in 1960, and the 1960, Alloy 600 was wonderful.
(Laughter.)
MEMBER ROSEN: If you do want to take my comment as a guide, I don't throw anything
away. So come look at my garage.
(Laughter.)
CO-CHAIRMAN SIEBER: You can have the head.
MR. SCHRAUDER: Is it seventeen feet in diameter?
(Laughter.)
MEMBER ROSEN: My garage you're talking about? Just about might fit.
MR. SCHRAUDER: This shows and addresses the issue on the containments. Yes,
we do have to cut both the Midland containment and the Davis-Besse containment
structure.
The Midland containment is a pre-stressed containment so it has to be de-tensioned,
and then we'll actually chip into that containment and open up a large, 20 foot
by 20 foot hole approximately.
The Davis-Besse containment is a shield building, a concrete shield building
with a free-standing pressure vessel and an annular region in between.
We are using Bechtel Power to assist us in the opening of both containments.
They have done most of the containment openings and restorations in the United
States.
We have a bullet on here that shows we will bring the head, the existing Davis-Besse
head out, protected and the people around radiologically from that.
Temporarily, we hope right now -- I should get analysis back next week that
will categorize what class waste it would be. And then it would be our intention
at this time to dispose of it if it is categorized as a low level waste rather
than create a temporary storage facility at Davis-Besse for it.
We will work with the NRC on that and make sure that the rest of the industry
knows that's our intent in case there is any desire to do any more examination
or testing on that head.
We are going to transfer our service structure from the Davis-Besse head to
the Midland head. We are putting in the inspection, the inspection modification.
That goes on the lower skirt, and that piece of the Midland head we will use,
and before we ship it to Davis-Besse, it will have that modification performed
on it to provide adequate inspection and cleaning of the head as necessary.
Of course, I have already said that we will re-use all of our control rod drive
mechanisms on this head.
As we were repairing the head and we had to cut out a couple of nozzles, we
had to reconfigure our core at Davis-Besse, specifically, the control rod locations
to assure that it was acceptable. We will go back to the original core design
with the new head. We will be submitting that core analysis to the NRC.
There are a couple of modifications that have been made over the years for serviceability
and outage flexibility, the nozzle flange split. Split dot ring modification
will be performed, and we will use the upgraded gasket design on the control
rod drive mechanism flanges.
MEMBER LEITCH: Have you thought about foreign object damage when you're cutting
a hole in the containment?
I guess you're planning to do this with the fuel still in the --
MR. SCHRAUDER: No, sir.
MEMBER LEITCH: You're going to de-fuel?
MR. SCHRAUDER: We will be full core off-load when we do the -- and I meant to
say that. The cut at the Davis-Besse site will not be a classic cut and chipping.
It will be a process that uses a very high pressure water lancing that essentially
washes the concrete off of the rebar, and Bechtel has used this process in Spain
several times.
There appears a much nicer cut on the containment and avoids having to chip
back to get the rebar exposed. Then the rebar can be tagged, cut, and then restored
right back into the original location so that it's already go the proper bend
to it and you cad weld it back in and then restore your concrete.
So it's a much gentler process.
MEMBER LEITCH: But even so, are you going to deck over some areas to prevent
foreign object outage or --
MR. SCHRAUDER: We do have a vessel cover for the Davis-Besse head. That will
be in place when we take the fuel out of the reactor vessel. Yeah, we'll be
very cognizant of foreign material.
They're spending a lot of time cleaning that containment up right now, too.
So I'm sure that it will be left very clean when we're done with it.
MEMBER LEITCH: One issue that always concerns me when you have a major construction
project like that going on. Its fire fighting capabilities, just I'm sure you're
going to get into a lot of detailed planning, but I would just like to remind
you to be sure that you haven't temporarily removed from service any of your
fire fighting capability while you're doing that because when that kind of activity
goes on, it just increases the potential for fire, and you want to be sure that,
you know, all of your fire fighting stuff is up to speed.
CO-CHAIRMAN SIEBER: Operable.
MEMBER LEITCH: Operable, or if not, some other temporary provision has been
made.
MR. SCHRAUDER: I agree.
The next slide, Slide 31, just shows some of the post installation inspections
that we'll do to verify that we do have a good fit on this. We'll fill and vent
the RCS, do a visual for gross leakage, and we'll bring the plant to normal
operating temperature and pressure with reactor coolant pump heat.
Of course, we won't be able to get right up into the nozzle space at that time.
So what we'll do is we'll bring it up to temperature and pressure. We'll cool
back down, and then we'll go in and look for visible signs of leakage when it
was at pressure.
We'll perform the control rod drop time testing in accordance with our tech
specs to verify the control rods do, in fact, go in at the appropriate speed.
Once we put the head on and you latch the control rods, you're pretty well satisfied
that you've got the proper alignment here, but we will, as required by tech
specs, do a control rod drop test.
The next page we don't really need to go into. They are approvals that we would
need from NRC staff, and the top two there were actually needed for our existing
head also in their IS program.
MR. POWERS: Okay. If there's no further questions, we'll turn it over to Steve
Loehlein to talk about the root cause updates.
CO-CHAIRMAN FORD: I just got a proposal here from Jack. Has everyone read the
root cause report?
It may be -- and I don't want to put you out of business.
(Laughter.)
CO-CHAIRMAN SIEBER: That was a god report, Steve. It really was.
CO-CHAIRMAN FORD: It was a very pointed and honest report, I thought.
Maybe the best way to tackle this in the cause of time is does anyone have any
questions having read the root cause report.
MEMBER APOSTOLAKIS: Well, maybe you can go to the inspector summary on Slide
52.
CO-CHAIRMAN FORD: Do you mind? Do you feel s as though you're being done out
of --
MR. LOEHLEIN: I don't mind. We thought that perhaps that time line slide would
have had some questions on it, but if people are familiar with that, having
read it, whatever is of interest to it, that's why we're here.
CO-CHAIRMAN FORD: It was a very complete report, I thought, and I enjoyed reading
it. I didn't enjoy it.
CO-CHAIRMAN SIEBER: I didn't enjoy reading it.
MR. LOEHLEIN: I didn't enjoy writing it all that much.
CO-CHAIRMAN SIEBER: But it was well done.
CO-CHAIRMAN FORD: Okay. Why don't you put the time line graph up just to jog
any people's memory as to whether this question --
MR. LOEHLEIN: It's probably -- I don't know by number. It's the fourth slide
in.
CO-CHAIRMAN FORD: It's this one here.
CO-CHAIRMAN SIEBER: We all have it separately.
CO-CHAIRMAN FORD: Why don't you walk through that one, and it might jog people's
memory as to the questions, and then go to the conclusion?
MR. LOEHLEIN: It's a little bit hard to do here logistically. So, Mark, I'll
ask you to go ahead and point.
this is a little bit of clarification on the way this is laid out. You start
at the very top of this diagram. We have a set of blocks that indicate what
we call industry and regulatory knowledge, milestones.
At about the 1995 time frame with the boric acid corrosion guide book, and I'll
pass on through, up through the bulletins and generic letters, and so forth.
As you proceed down, the first thing you see is is a blue bar graph. The blue
bar graph indicates the reactor coolant system and unidentified leak rate over
time.
There is also the red dashed line that proceeds on a diagonal from left to right
with three data points on it or the number of nozzles that were not visible
in an as found state, those refueling outages.
As you continue on down this chart, you run into the yellow colored blocks that
indicate the containment radiation monitor filters and the change in preventive
maintenance frequencies brought about by clogging either to boric acid or to
a combination of boric acid and iron oxide.
Below those blocks we have similar blocks reporting the frequency of containment
air cooler cleanings, and beneath those, we have the two time lines. The first
one is simply the chronological passing of years. Beneath that are the outages
and plant cycles as they line up.
Then in the numerous blocks down below, there's actually three sets of data.
As you read from left to right, the first set of blocks is the conditions for
the control rod drive mechanism flanges.
The next set below it is the reactor pressure vessel flange itself on the outside
perimeter, and then the bottom set of blocks is the reactor pressure vessel
head.
So that's how this is laid out. Any particular questions on it?
CO-CHAIRMAN FORD: I've just got a generic question. I must admit I read it in
anticipation of reading -- because of my interpretation what a root cause report
is -- that it would tell me specifically what the mechanism was and thereby
when things started, and that would give me some idea as to how generic this
was and whether it was a leader of the fleet.
And of course, it didn't have that, but having heard the reports earlier on
from NRP, I'm assuming that that onus is now being passed to the NRP; is that
correct?
MR. LOEHLEIN: I think --
CO-CHAIRMAN FORD: They will take on the burden of determining whether this really
is --
MR. LOEHLEIN: We probably each have a piece in that answer. So I'll speak first
and say that clearly in the evidence we had available to us in the large cavity
region at nozzle three, we could from the plant data and other physical evidence
say pretty much what happened since about 1998.
But that only describes what happens at high corrosion rates once the conditions
are right, boric acid and so forth.
And what we all know and what we need to study further is what happens prior
to that, and we didn't have measured data that we could go to and say how long
the steps took, and that's the kind of work I think Christine at EPRI is taking
on.
MS. KING: Right. This is Christine King with EPRI.
We took that on, as Glenn said earlier, just a couple of weeks after the discovery
of the wastage at Davis-Besse because of the idea of understanding how this
progresses, and we will obviously continue to work on that.
CO-CHAIRMAN FORD: The thing that keeps coming to mind, everything from stress
corrosion cracking of turbines to tracking of small pipes: big pipes will never
crack, and sure enough they do crack.
In Japan, we will never crack a pipe in Japan. And they do.
And so whenever anyone says that this is a one off (phonetic) situation, my
ears immediately start to prickle, and my hair starts to prickle.
But anyway, I'm really suspicious until we understand what the real root cause
was and how it relates to geometry and chemistry, et cetera. And this is why
I was urging you to as quickly as possible we'd better put this one to bed.
MR. LOEHLEIN: What I would comment on is in all this investigation, we did as
a team with a technical experts and so forth, is that we were unable to uncover
any new evidence to provide us with any kind of insight different from what
is already known, and that is that cracks can lead to leaks, can lead to corrosion
if it's not discovered.
CO-CHAIRMAN FORD: And one of the conjoint requirements to have.
MR. LOEHLEIN: Or detected.
MEMBER BONACA: I would like to -- we're talking about root cause, and so your
conclusion is that inadequate inspections of the closure head was the problem.
I think beyond that it seems to me that the fact that you cannot fix the flange
leaking completely at any given outage, but you manage that issue by saying
we will fix the most severely leaking and we'll leave the rest must -- everything
from that point on, in fact, you concluded that, you know, presumed boric acid
leakage was coming from the flange, and so you kept doing that.
And then you presume that the accumulation of boric acid crystals on the head
was coming from the flange. Therefore, you kept managing the issue, and that
prevented you from performing complete inspection.
So I'm saying that to me the lesson learned is that when you have an issue of
that kind you do not manage it. You just simply fix the flange leakage so you
don't have it anymore. Because otherwise it will have a cascading effect, and
your people are going to still live with a limited amount of time to perform
the fixing of those flanges, and that cascades in not having enough adequate
inspections.
I mean it seems to me that is throughout the root cause. There is that threat
that people wanted to do the right job, but they said, "Well, we've reached
the time limit. We could only fix this many flanges. So we'll leave this flange
for the next outage."
MR. POWERS: Right, and I would say that there's a number of things in the root
cause that are beyond the technical root cause that we've discussed thus far,
and we're still ongoing with the management root cause issues. We're taking
actions at the site as a consequence of that.
MEMBER BONACA: Yeah, and I don't want to get inside that. I'm only -- when I
look at that and it says inadequate inspections, I think more than that is what
was the cause of that. I mean, in part it was because you really believed that
the leakage was coming from somewhere where you thought you knew, and that led
you to convincingly believe that you didn't need to inspect further because
you knew where it was coming from.
MR. POWERS: Right, and there's elements of problem solving adequacy.
MEMBER BONACA: I agree.
MR. POWERS: How far we drill down, and so we've got a number of things on our
list of things to do as part of our 0350 restart.
MEMBER BONACA: Yeah, because inadequate inspection could be interpreted as simply,
you know, we didn't look enough or whatever, but really there was this issue
fundamentally that we know where it's coming from. We don't have to look further,
and therefore, we can manage it. We can keep, you know, from outage to outage,
to push further fixing to the next outage.
And that seems like a threat that finally convinced a lot of your people at
the working level that that was the solution, and they kept believing it.
CO-CHAIRMAN FORD: If there's no more questions on the root cause aspect, I thank
you very much indeed, and thank you for coming.
I'd like to move on for the NRC.
Do you want a break? Okay. Ten minutes. We don't want any accidents. We'll recess
until ten minutes past five.
(Whereupon, the foregoing matter went off the record at 4:59 p.m. and went back
on the record at 5:11 p.m.)
CO-CHAIRMAN FORD: Okay. Thank you very much, Jim. I appreciate your giving us
the time.
MS. WESTON: This is a part of the NRC package, part of it.
MR. GROBE: Okay. We've got three more topics that the staff will present. I'll
update you on what we've been doing with respect to regulatory oversight at
the Davis-Besse plant.
Ed Hackett is going to be talking about independent lessons learned task force
that's been chartered by Bill Travers, and Allen Hiser is going to talk a little
bit about management by leakage detection.
I'm sure you're going to have no questions for myself and Ed and about 300 questions
for Allen. Are you ready for the next slide, Allen?
Allen is going to flip slides for me.
Just a brief time line of major activities that have occurred. Of course, the
pressure vessel had degradation on the 6th of March. The AIT inspection on March
12th, began on March 12th. We issued a confirmatory action letter on the 13th
an established the oversight panel on April 29th.
The basis for chartering an 0350 panel for Davis-Besse were fourfold. First,
the situation at Davis-Besse represented a significant, complex technical issue
and also a complex regulatory issue.
The plant is in an extended shutdown and regulatory hold, in effect, and that's
the confirmatory action letter.
The 0350 panel would enhance the agency's focus on clearly defining and communicating
the plant specific issues that need to be resolved prior to restart, and we
provide as a panel a focused and coordinated oversight.
The next slide is -- please stop me if you have any questions. I'm just going
to zip through this -- goals of the panel are several. One of the goals is that
the panel provides oversight and assessment of licensee performance. It's a
broad and integrated focus on assessment, much more comprehensive than would
be applied to a routinely operated plant.
We assure that the restart issues are identified and resolved, and what's critical
here is a shared understanding between First Energy, the NRC, and the public
on what those issues are needing resolution prior to restart.
We have the capability to coordinate across organizational boundaries within
the agency, and of course, Region III, NRR Research, Public Affairs, Congressional
Affairs, ACRS. There's been many aspects of the agency that have been involved
in the Davis-Besse issue.
Provide a single point of contact, a single focus for communicating with external
stakeholders. We've had extensive interface with concerned citizens in the area,
concerned groups across the country, federal, state, and local elected officials,
and of course, the media.
So it's important to have a single focus and a cohesive message on what's going
on at Davis-Besse.
MEMBER LEITCH: John, it's my understanding the 0350 panel goes on through, I
guess full power operation.
MR. GROBE: Yeah, I'll get into that in a little bit more detail.
MEMBER LEITCH: But as far as identifying restart issues, other than the obvious
replacing the head, is there some kind of a report or a point in time when those
restart issues are clearly defined?
MR. GROBE: Yes.
MEMBER LEITCH: And what is that point in time?
MR. GROBE: There's two documents that guide the activities of the 0350 panel.
One is called process plan. That's been promulgated and issued publicly, and
it covers more not plant specific per se, but process issues, including interfaces
and communications and activities that need to be accomplished.
The second document is called a restart checklist, and that is the document
where those specific issues that need resolution prior to restart will be clearly
communicated. A checklist has not been issued yet primarily since the licensee,
First Energy, has not completely defined the causal factors in some of the areas,
and I'll get into that in a little bit more detail in a minute.
MEMBER ROSEN: What was the first document's name?
MR. GROBE: Process plan.
MEMBER ROSEN: And that is on the Web site?
MR. GROBE: Yes, it us.
CO-CHAIRMAN SIEBER: It's in the inspection manual chapter, 350.
MR. GROBE: Right. There's guidance in the manual chapter, and you interpret
the guidance that's in the manual chapter as applied to the specific task. Each
plant that might come into an 0350 might have different characteristics required.
MEMBER ROSEN: So if I go to the process plan, I'll see the actual milestone
dates for Davis-Besse?
MR. GROBE: No, no. There are no dates.
CO-CHAIRMAN SIEBER: You'll see general format.
MEMBER ROSEN: That's what I was still interested in. Is that what you were asking
about, Graham? What the dates were for when we would see --
MEMBER LEITCH: That's what I was asking about. I think I heard that the dates
are not yet established.
CO-CHAIRMAN SIEBER: Right.
MR. GROBE: We won't establish --
CO-CHAIRMAN SIEBER: The issues aren't established.
MEMBER LEITCH: But the process plan is not specific to Davis-Besse. It's more
or less a checklist of those things that one must consider --
MR. GROBE: Right.
MEMBER LEITCH: -- before moving to restart.
MR. GROBE: We will serve no wine before its time.
(Laughter.)
MR. GROBE: You won't find dates in our documents. Like I said, we will develop
a shared understanding of those issues that we expect to be resolved prior to
restart.
When the licensee believes that each of those is ready for evaluation, we will
provide inspections of those activities and then address any findings with the
licensee.
So there won't be any dates in our restart plan, our process plan.
The panel provides continued oversight after plant restart. Our expectation
is that the panel will continue to provide that oversight at Davis-Besse for
at least one calendar quarter following restart.
And finally, we create copious amounts of documentation. All of our internal
meetings and external meetings are documented, and those are available on the
Web site.
We're now going to be transcribing the meetings that occur in Ohio to make sure
that people who can't make it to Ohio have access to the specific issues that
are discussed.
The panel members include two senior managers, one from Region III, myself,
and one from NRR; three supervisors, two from Region III and one from NRR: the
NRR project manager; the senior resident inspector; and a risk analyst from
my staff in Region III.
So as I said before, it's a very broad oversight. It brings together a variety
of different skills from different parts of the agency.
The routine reactor oversight process, what's come to be known as the ROP, is
suspended in the situation where you have a plant that goes under 0350. There's
a number of reasons for that.
One is that the plant is in a configuration that the reactor oversight process
was not written to address.
In addition to that, a variety of the operationally focused performance indicators
will atrophy when the plant is shut down. So those PIs will not be providing
insight into plant performance.
We talked about the process plan. The process plan will include coordination,
communication activities, inspection and assessment activities, licensing activities,
and a variety of things. It's about a ten-page document.
The restart checklist has not been issued yet, but that will contain all of
the restart items.
We have been averaging about two internal meetings per week, and we had our
first public meeting in early May. Our second public meeting at the site, in
the vicinity of the site is next Wednesday, a week from today.
The licensee has submitted what they call a return to service plan. That was
submitted on May 21st. That's also available on the Web site. There's what they
call building blocks. Is that -- yeah, okay. I'm getting nods back there.
Six substantive building blocks that need to be completed to return to service
effectively. Three of them are pretty straightforward. Three are a bit more
complex.
The reactor head resolution is a fairly straightforward activity, much more
straightforward now that the head is being replaced instead of repaired.
Containment extended condition, that includes extensive inspection of the reactor
pressure boundary, as well as inspection of other equipment inside containment
for damage or the effects of the environment that the equipment was subjected
to.
The other one that is pretty straightforward is the last one, restart and post
restart test plan. Those are fairly clearly understandable and definable activities.
The remaining three are a bit more complex in defining exactly what is necessary
prior to restart, the scope and depth of those activities. The licensee has
defined a system health plan where they're going to select risk significant
systems and evaluate those at some level of depth to insure that they actually
have what they thought they had as far as safety system health.
A program or a process review plan, where they're going to pick at least three
programs, I believe: the boric acid management program, of course; the corrective
action program; and the design change program, and possibly others that they're
going to review at some level of detail.
And the next one is one that has not yet been fully developed yet, and that
is the management and human performance excellence plan. There's been, I think,
four different activities that have been undertaken to try to get their arms
around exactly what went wrong from an organizational effectiveness perspective,
a human performance perspective, management effectiveness.
That included a group chartered by INPO, which was senior executives from a
number of plants that came in an evaluated what happened; a group chartered
by Bob Saunders, the Chief Nuclear Officer, that included review of various
activities; the root cause team, of course; and there was one other. It slips
my mind at the moment.
But the licensee is now accumulating all of that data and is going to define
what it believes are necessary activities prior to restart.
Not only is it difficult to understand the scope of what activities in these
areas are necessary, but how to measure success is not an easily defined concept.
So those are the areas where we're going to be having some dialogue in our public
meetings at the site.
MEMBER APOSTOLAKIS: Do you have any guidance as to what a human performance
excellence plan is?
MR. GROBE: No. The way I've approached these kinds of activities in the past
is really four steps. First is insuring that we have confidence that the licensee's
identification of causal factors is sufficient.
Second, to insure that the scope of what their activities that they're going
to undertake -- they define these activities, and we make sure that the scope
is sufficient to address the root causes, the causal factors.
We'll provide inspections of their implementation of that plan and then resolve
any deficiencies, and there could be a substantial number of deficiencies that
we identify that don't need to be resolved prior to restart that can be ongoing
activities after restart.
But there is no specific guidance in that area. Clearly there's a number of
performance, organizational effectiveness and performance issues that contributed
to what happened at Davis-Besse. So we'll be making sure that they identify
those to our satisfaction and that they have a plan to assess how they're improving
in those areas.
There's three inspections that are ongoing right now: the AIT follow-up. The
primary focus of that is taking the results of the AID inspection and putting
them into a regulatory framework, what are violations, what aren't violations.
There are some technical issues that have come out of the AIT that we'll be
forwarding on to headquarters for evaluation.
The head replacement plan we've received from the licensee the process that
they're going through and milestones, activities so that we can start scheduling
our inspection activities, and the extent of condition inside containment inspection
has been initiated.
Those are the activities that I wanted to cover with respect to what we're doing
at Davis-Besse today. There were two issues that came up earlier in the day
that I wanted to comment on.
One, Dr. Apostolakis, you raised an issue regarding the resident inspector knowledge
of the head inspections. The resident's primary focus is on operational safety,
day-to-day operational safety, and that encompasses operator performance, equipment
operability, maintenance activities, testing activities. It's at least a full-time
job for the two residents that are on site.
We're rather protective of distracting their focus off of operational safety.
For PWR linguists, you're at risk of losing the bubble if you distract the residents
from their operational safety focus.
Members from my staff, particularly several metallurgists, would be the ones
who would be going out to the site and observing the head inspections that licensees
have undertaken.
The challenge with that is that obviously they're traveling out of the regional
office. So they can't be everywhere all the time. We have to depend, as Bill
Bateman mentioned earlier, on the veracity of the statements made by the licensee,
and we challenge those through phone calls and the residents participate in
that, and they have some awareness of what the licensee has been doing.
But I wouldn't expect them to get into detailed evaluation of the head inspections
because it would take them away from their principal responsibilities.
MEMBER APOSTOLAKIS: Let's see. I mean, the fact that the containment filters
had to be replaced much more frequently than originally anticipated, isn't that
something that somebody ought to notice?
MR. GROBE: As soon as that issue came up, I know we in Region III assessed that,
and the containment air cooler cleaning and the red monitor filters, and the
resident inspectors did that.
And of course, the information notice was issued. So the licensees were also
sensitized to that. So we did follow up on those types of indicators and found
no problems at the other sites in Region III.
MEMBER APOSTOLAKIS: No, I mean at Davis-Besse.
MR. GROBE: Oh, in retrospect?
MEMBER APOSTOLAKIS: Yeah.
MR. GROBE: There were two inspections in the fall of 2001, and the resident
inspector had become aware of operational concerns with the -- this is actually
a leakage detection system, the RAD monitors, and focused both on the operational
performance of that system, as well as the source of the corrosion.
The licensee had committed at that time to do a comprehensive inspection. They
did do some evaluation in containment of sources of leakage, but did not identify
any and committed at that time to do a comprehensive assessment during the 2000
outage.
I misspoke. It was the fall of '99, and so they committed in the 2000 outage
to do a comprehensive evaluation of what might have been leaking in containment.
In fact, that's one of the issues that Ed Hackett's group is going to be looking
at, is how we followed up on that organizationally; the inspection program,
how it addresses issues of that nature.
MEMBER APOSTOLAKIS: One last question.
MR. GROBE: Sure.
MEMBER APOSTOLAKIS: It's really a comment. When we were discussing with the
staff the revised oversight process, this committee expressed concern about
the safety conscious work environment cross-cutting issue, and the issue that
we raised was, you know, how are you going to know that the safety conscious
environment is, in fact, acceptable.
And the answer was: we're not going to do much about it because if it is not
good, we're going to see it in the hardware. Things will start failing or, you
know, doing things.
I wonder now as a result of this experience whether we still believe that that's
the case, and do you?
MR. GROBE: Again, that's an issue. I had the distinct pleasure of spending four
hours with the lessons learned task force yesterday, and that's an issue that
they're going to be asking.
The results of our inspections and PIs and assessments over the last really
decade or more of Davis-Besse performance has shown good performance. We do
inspect the effectiveness of their corrective action program, and that gets
to a certain extent to this safety conscious work environment or safety focus
of the folks at the facility, and those inspection results revealed the program
was operating effectively.
MEMBER APOSTOLAKIS: So in retrospect then, we have to rethinking that.
MR. GROBE: That's correct. We have to look at what lessons we can learn, and
that's why --
MEMBER APOSTOLAKIS: Now, I don't know if you want to make a comment on it, but
I believe the problem is that this agency does not have the tools to do that.
Now, you may not agree with me, but --
MR. HACKETT: I think I'd add the comment. I think Allen is going to get into
this. One of the early themes, if you can call it a theme, in the lessons learned
task force is let's look at management of these issues through leakage, basically
through leakage management, and obviously in this case, you know, you've eroded
margins to the point there is effectively no margin.
And that does go back to what tools are available to do better than that because
in several instances now we've gotten to these points by people finding leakage,
either NRC or in most cases licensee inspectors, and it's going to challenging
the adequacy of that and then how do you do better.
You can do nondestructive examinations, but they're costly. They may not be
entirely effective at going after exactly what you're looking for. So I think
it does go to development of the tools, and I think that's going to be one of
the things to come out of it.
MEMBER APOSTOLAKIS: Good.
MEMBER LEITCH: I know we don't want to go too far down that road, but that inspection
of the corrective action program is not an ongoing inspection. It's module 4500,
right, which is done every two years or something like that?
MR. GROBE: It's got a new number today, but, yes, it used to be 4500.
MEMBER LEITCH: Yeah, right, and so someone comes in from the region and looks,
I guess, retrospectively at the effectiveness of the corrective action program.
MR. GROBE: The assessment of the corrective action program is in three phases
today. The first part is a certain portion of each inspection procedure, each
inspector every time they go out whether it's a health physics inspector, an
engineer, resident inspector, a certain portion of their time during each inspection
is focused on selecting certain activities retrospectively and making sure that
those activities were properly resolved. So that's one part.
The second portion is that we just recently changed the periodicity of the major
inspections from annually to once every two years. The reason for that was that
freed up a number of resources.
It did two things. It gave us more time when we do it once every two years.
We added about 25 percent to the duration of that inspection. So it gave us
more time and more resources when we actually do go out to get more intrusive.
Secondly, it freed up a number of hours to select certain activities that are
ongoing during that two-year time period between inspections and really drill
down deeply. The more complex issues that come up, we can go out in a more real
time basis and send an inspector out or the resident can do these kinds of inspections.
So it's in those three phases. We have a major team inspection every two years
where it's a risk focused selection of quite a few deficiencies that have occurred
over the last two years and evaluating how they resolved those; the real time
situation between two years where we drill down and every inspector every time
they go out samples.
MEMBER BONACA: But it seems to me, following up on this issue, oftentimes we
see this concern with inspections, adequacy of inspections, and all of the ROPs
focused on performance of safety systems, which are really managed and maintained
on line outside of the outage.
And it seems to me that an area of concern would be to look at the outages specifically
because there you see the constraints of activity, length of time given to activities
that leads to inadequate corrective actions, inadequate inspection, and so on
and so forth.
And that really is what is more likely to have a conflict between the need to
restart and taking care of business completely. So I know you do have, in fact,
your active inspections during outages, but is it -- I think still you have
the resident inspector simply there just looking at what's going on, I mean.
Are there any lessons learned there? And should it be stepped up, the focus?
MR. GROBE: I bent the lessons learned task force here on a number of these issues
yesterday. In today's environment, competitive environment, outages have been
getting shorter and shorter, and outages are frequently less than 20 days now.
It becomes more and more difficult for us to inspect those kinds of activities
that are only available during outages. So that's a challenge for us.
We try not to schedule complex inspections during outages because the entire
work force of the facility is focused on the outage. So we try not to distract
them from that focus.
So it's a challenge, and that's one of the issues that is before the lessons
learned task force.
MEMBER BONACA: That's the major trend in the industry performance, has been
the shifting towards shorter and shorter and shorter outages, moving out, for
example, you know, all of the maintenance equipment, all line, when it's done
without the pressure of the outages. So, therefore, you have much higher assurance
that the work will be done properly.
And so it seems to me that there would have to be almost like a revisiting of
the focus on that outage because that outage becomes critical, and the pressure
in on the operators. I mean, I know I've spoken with some of them, and they
have told me they feel the pressure from peers, who are really competing with
them, and then from their management because if somebody else is doing it shorter
and shorter time, why not us?
So, you know, I think certainly that's an area where I understand it's a challenge
for you, but you know, one may conceive that you would want to have less focus
at large on those activities which you know have been dedicated resources and
time like staff under maintenance rule and more conceptive (phonetic) teams
maybe, you know, just focusing on outages.
MR. GROBE: One of the things that we've observed as outages have gotten shorter,
of course, as you mentioned, some activities have been taken out of outages
and put on line, but one of the other things that we've observed is much more
complex and effective scheduling and work management activity, which actually
improves the quality of work.
There is that additional schedule pressure, and we're sensitive to that, but
in fact, we've seen the outages are better managed, and that's one of the ways
that the outage schedule has gotten compressed.
MEMBER BONACA: Well, no, I agree. I mean, they can do it. If they haven't done
a very strong improvement affecting the way the outages are managed, then there
are a lot of things.
However, time pressure is still time pressure. There are going to be some things
which are a decision is going to be made that this is not important enough that
we have to do it completely or this can be postponed, whatever. It has to be
done because time is more limited.
MR. GROBE: And that's, quite frankly, one of the issues that is part of our
follow-up activities at the AIT, is looking at those specific questions.
MEMBER ROSEN: I'd like to come away from the discussion of the outage for a
minute and come back to your earlier remarks about operational focus, which
I absolutely commend. I think that is the right thing for the inspectors to
do, but I'm puzzled by that comment and the fact that what was going on at Davis-Besse
for perhaps four years or maybe more was an event, an ongoing event, of the
degradation of the head which sent a lot of signals, operational signals, the
containment atmosphere, coolers, pressure drop, and the need for recurrent cleaning
of that.
Just take that for an example.
MR. GROBE: Sure.
MEMBER ROSEN: There's clearly an operational event that your inspectors with
their operational focus had to know about and had to draw a conclusion about.
MR. GROBE: In fact, I'm not sure that we had focused on the containment air
cooler cleanings. I just don't think it rose to the level of cognizance on the
residence staff, and Ed Hackett and the rest of the lessons learned task force
team will be out interviewing all of the inspectors, but for my interaction
with them, I don't believe that came to our attention.
MEMBER ROSEN: Well, clearly, in hindsight, which is always 20-20, one would
say that that was maybe the preeminent signal to inspectors who have an operational
focus that there was something amiss.
MR. GROBE: I think that's clearly one of the signals. The other one is the RAD
monitors, which was probably more directly connected to what was going on. I
believe it was July of '99 that they sent the sample filter out to be analyzed,
and it came back that there were corrosion products that were produced in a
steam environment.
That was a clear message that there was some leakage going on, primary coolant
system leakage, and that did come to the attention of the inspectors through
their routine inspections, and they did follow up on it, and it's documented
in two reports.
It didn't get above the resident supervisor, and it didn't come to the cognizance
of myself or the division reactor projects director.
We asked the right questions, but maybe didn't follow up the way we should have.
CO-CHAIRMAN FORD: I'd like to move on if I may.
MR. GROBE: Sure.
CO-CHAIRMAN FORD: Because this is not a topic that we covered in the letter.
MR. GROBE: There was one other if I could
CO-CHAIRMAN FORD: I'm sorry.
MR. GROBE: There was one other question that came up, and I just wanted to make
sure despite Research's desire to be done with the finite element analysis.
That really is an important activity for two reasons, and I think they kind
of came up, but I just wanted to make sure. One of the things that is part of
the new program is a new definition of how we communicate significance to the
public, and the results of that analysis and the following analysis, which will
be the probablistic assessment. That will feed the probablistic assessment and
are critical to us in our ability to communicate the significance of this event
both internally and to the public.
The second though is we also use the results of that analysis to budget staff,
and the more significant the finding, the more staff we put on a project.
And one of the things that I also bent the lessons learned task force's ear
yesterday on was, you know, we've shifted to a, quote, risk informed framework.
The significance determination process is actually risk driven in this arena.
In other areas like health physics and emergency preparedness and security,
it's risk informed.
But in the areas where we can do probablistic analysis, it's fairly well risk
driven. You heard some analyses both from our Office of Research, as well as
the licensee's staff, on burst pressure of the remaining cladding. It will be
interesting to see how that's handled within the significance determination
process and, when we're done with that, whether that truly reflects the significance
of the performance deficiencies.
And that may be an opportunity to reexamine the way we do risk significance
and whether there should be some other factors that are considered.
Taking notes, Ed?
MR. HACKETT: In fact, I am.
MR. GROBE: Good. Those were the other issues.
CO-CHAIRMAN SIEBER: So it's going to be green.
MR. GROBE: If you looked at it as a binary gate, you could come to that conclusion,
but, in fact, there's probability distributions on all of those things. So even
though the burst pressure might be some psi, that doesn't mean it wouldn't fail
at a lower pressure.
CO-CHAIRMAN SIEBER: Right.
CO-CHAIRMAN FORD: Ed, thank you very much.
MR. HACKETT: I think like Jack said, Jack had already reached several conclusions
for the task group yesterday.
MR. GROBE: If you need any help, just let me know.
MR. HACKETT: I think we're going to get all kinds of help.
I guess given everything that's been discussed here and the situation, it's
not surprising that we're talking about a lessons learned task force. The agency
has done these before. We don't have criteria for deciding exactly when they
might be done.
The last one was done for the Indian Point Unit 2 tube rupture; this one for
Davis-Besse reactor vessel head degradation.
I'm the assistant team leader. Art Howell from Region IV, he's the division
director, division reactor projects in Region IV -- division reactor safety.
I'm sorry.
MEMBER LEITCH: Who is learning the lessons here? In other words, is this an
internal --
CO-CHAIRMAN SIEBER: If anybody?
MEMBER LEITCH: Is it the NRC going to look at Davis-Besse or look at the NRC's
performance?
MR. HACKETT: I'll make several comments in that regard. I guess go ahead and
put up the next slide here to get into some of that.
The primary focus, as you are indicating, is on the NRC and the NRC's internal
processes. It's not limited to that though, however. It's also to look at recommended
areas of improvement, both the NRC and the industry.
We also say reactor vessel head degradation. The scope and charter is actually
broader than that. I think you can use --
MEMBER APOSTOLAKIS: I'm really confused. If it's broader, why doesn't it say
that? Why do you have to say, "But really it is broader"?
It always confuses me.
MR. HACKETT: It was written before I got there.
(Laughter.)
MR. HACKETT: So I guess the charter --
MEMBER APOSTOLAKIS: Because that was my next question. Why limit yourself to
reactor vessel --
MR. HACKETT: That's a good question. It was written this way. I think the charter
is publicly available now on the NRC's Web site, and if you go below this basic
mission statement, it does say that it is to consider other areas, you know,
basically.
Especially in this case, looking at reactor coolant pressure boundary leakage
in general, you know, would be more consistent with the charter.
MEMBER APOSTOLAKIS: I would not defense in depth the scope of the task force.
MR. HACKETT: That's a good point, too.
The other point I'll make, since we're literally just kicking this thing off
this week, we are looking for public comment, soliciting public comments on
the charter. I'll get into the charter here in a few minutes.
So far we have a charter that's been written. That was written before the team
was even in place, and the charter is still open to suggestion, comment from
the committee, from the public and others.
MEMBER APOSTOLAKIS: Let me understand something. If I go -- I haven't done it;
I should do it -- if I go to the NRC Web site and look up reactor oversight
process, Davis-Besse, am I going to see greens all over the place?
MR. GROBE: yes.
MR. HACKETT: I believe so.
CO-CHAIRMAN SIEBER: I told you.
MEMBER APOSTOLAKIS: Huh?
CO-CHAIRMAN SIEBER: I told you.
MR. HACKETT: Yes.
MEMBER APOSTOLAKIS: I believe you.
MEMBER LEITCH: For the last two assessment cycles.
MEMBER APOSTOLAKIS: Okay. So there must be some lessons learned.
MR. HACKETT: I think there will be some.
MEMBER APOSTOLAKIS: There will be some. Okay.
MR. HACKETT: Maybe a couple other things I'll mention up front here in terms
of coordination and interfaces. There are other investigations going on that
I'm sure the committee is aware of and others are aware of.
The Congress, Energy and Commerce Subcommittee, I believe, has an investigation
ongoing. I believe they've been out to the site. They will likely be talking
to the NRC, probably to the lessons learned task force, to Jack and 0350, and
there are others.
There's Jack's 0350 panel, obviously. The Inspector General, internal to the
NRC, is also looking at the NRC decision process leading up specifically to
delaying the inspection at Davis-Besse.
So those are going on. Those are going on in parallel with this.
MEMBER APOSTOLAKIS: Would it be appropriate to add safety conscious work environment
there?
MR. HACKETT: That is part of what we'll be looking into.
MEMBER APOSTOLAKIS: Of the oversight process, yes.
MR. HACKETT: So yes.
MR. GROBE: They asked me many questions yesterday about the corrective action
program inspections and about inspection perform --
MEMBER APOSTOLAKIS: Who's "they"?
MR. GROBE: The task force.
MEMBER APOSTOLAKIS: Oh, these guys?
MR. GROBE: Yeah. They were brutal.
CO-CHAIRMAN SIEBER: Well, there's one thing about examining the corrective action
program, and that's if the standards are low enough and there's not a questioning
attitude. Then there's not much in the program, but everything that's in there
probably gets corrected.
And so that's part of it, which the inspection maybe doesn't get to.
MEMBER APOSTOLAKIS: These guys will define questioning attitude every six months.
He will come back and say that the definition is this, right?
MR. HACKETT: I wish we had six months.
MEMBER APOSTOLAKIS: We all talk about it, but we don't know what it is really.
CO-CHAIRMAN FORD: Well, the ACRS certainly has it.
(Laughter.)
CO-CHAIRMAN SIEBER: The question would be what's all of that red stuff coming
out of that hole.
MEMBER APOSTOLAKIS: And the answer would be: don't worry about it.
(Laughter.)
CO-CHAIRMAN SIEBER: That's standard.
MR. HACKETT: We'll come to the schedule in a bit, and I'll wish I had six months,
I'm sure. Actually it's mandated to be done in about three months, almost exactly
three months from today. So it's an ambitious effort.
The charter elements are listed here as we have them right now. There's really
these five pieces with an awful lot of the front end focus is going to be on
the reactor oversight process, and I think Jack covered that more than adequately.
In addition to that, regulatory process issues at the NRC, including evaluation
of the regulations, licensing review process, regulatory processes, such as
the generic communications and the clarity thereof for regulatory process.
An element on research activities. We've heard from the Research Office today,
and that's my home base. So there are obvious issues with not just the research.
This isn't restricted to the NRC Research Office. This is research activities
in general.
Should there have been some things that should have been being done that might
have led us to be in a better place to identify this type of thing from a research
perspective or to mitigate it more successfully?
So we'll be looking at that type of thing, including research performed external
to the NRC.
International practices. I think it's pretty obvious that some of the foreign
industry has looked at this issue very differently than the United States did.
Most aggressively handled in France, and I think Allen has presented this many
times to the committee.
With the initial discovery of Bouget in 1989, they embarked very quickly thereafter
on a head replacement program, which, you know, we didn't do after discovery
of some axial type indication in maybe like the '97 time frame, general letter
9701.
At any rate, it has been handled differently for some very different reasons,
but the lessons learned task force will be looking into that, too.
The generic issue process, there have not been generic issues associated with
boric acid corrosion or much involved with corrosion in general. That will be
one of the topics.
Should there have been? Should there be now? Should this process somehow be
better tuned to picking these kind of things up? Because that type of thing
has not happened.
So at least those five elements are there. One of the things I'll mention right
now is the EDO feels strongly about soliciting input on this charter. So I'd
be glad to take input that anyone might have.
MEMBER APOSTOLAKIS: Yeah, are we going to see you before you publish your results?
MR. HACKETT: Well, I guess that's probably largely up to you guys. We're going
to be plenty busy enough. So I guess I didn't come here, especially from Art's
perspective, to be volunteering too many presentations over the three-month
period.
I would think what I'll come to in some of the subsequent slides here is that
we have a period where we're basically in a preparation phase right now. We've
literally just assembled a team this week.
The review phase really starts at the end of June and should complete more like
the end of July, and by then there will probably be -- there will be a developing
story, obviously, along the way, but by then there would be something to tell,
and we would be in the mode of trying to integrate it and writing the report
at that point.
So that might be a point to talk some more about. It will be briefing. Obviously
internally we report directly to the Deputy EDO, Bill Kane, and to the EDO,
Bill Travers. They'll be receiving at least weekly briefs on the progress of
the task force.
And if the committee would like to hear, you know, an update --
MEMBER APOSTOLAKIS: I think we need to discuss that in private.
MR. HACKETT: That's something we'll take as an action.
MEMBER ROSEN: We have a discussion of the schedule for this weekend.
MEMBER APOSTOLAKIS: No, but this is something new.
MEMBER ROSEN: Right, but I think we can take this up.
MEMBER APOSTOLAKIS: Yeah, yeah.
MEMBER ROSEN: I'm saying on Saturday.
MEMBER APOSTOLAKIS: Sure.
MEMBER WALLIS: This second bullet, does that include looking at how we might
view risk informed regulation as a result of what we've learned?
MR. HACKETT: I think that's fair. That one is fairly broad in terms of regulatory
process. Certainly the NRC processes have been focused at performance based
risk informing for a number of years now. So I think that's fair game under
that element.
MEMBER WALLIS: This kind of event isn't in the PRA, I understand, or is it?
MR. HACKETT: I don't --
MEMBER WALLIS: There's no analysis of --
MR. HACKETT: This specific event I don't believe would have been anticipated
to be in a PRA. I would think the -- I'll defer to Steve or others to answer
that more definitively.
I think what is or what has been evaluated, I know, is the LOCA that would result
from multiple rod ejection has been, and that was shown in terms of the LOCA
situation to be bounded by the hot leg break.
MEMBER ROSEN: PRAs typically don't address passive components. The head of the
vessel is a passive component. So it wouldn't show up as the component.
MEMBER WALLIS: Passive component about to become active.
MEMBER ROSEN: That's been fairly accurate.
MEMBER KRESS: LOCAs are all passive.
MR. LONG: This is Steve Long with NRR staff.
The PRAs typically address initiating events that would be failure as a passive
component to pipe break or whatever. So there's a medium LOCA frequency. There's
a small LOCA frequency, except for special initiators where you have postulated
a mechanism and gone in and analyzed the failures that lead to that mechanism,
such as an interface systems LOCA or something, you really just lump everything
that might create a hole of this size into an initiating event frequency.
MEMBER APOSTOLAKIS: We had recommended when we reviewed Athena that the project
look at the possibility of having an initiating event due to human actions during
normal operations.
You know, so before you go to the PRA, you have to do all this. Athena has to
take care of it, and then eventually, of course.
But you're right. Right now it doesn't have it, but these are -- I think the
problem is broader. I think there has been reluctance to get into organizational
issues, you know, for a number of reasons for the last several years, and these
naturally involve organizational issues, I mean, however you want to --
MEMBER WALLIS: You can fall back on Defense in--
MEMBER APOSTOLAKIS: Well, that's what I'm going to do, the structuralist approach.
What if you're wrong?
MR. HACKETT: I think you're --
MEMBER APOSTOLAKIS: Well, there has to be a way out of it, Steve. Either we
have to understand it or we put Defense in Depth, right? That's what Defense
in Depth does. It helps you when you don't understand.
CO-CHAIRMAN FORD: I'd like us to move on if we may.
MR. HACKETT: We're fortunate that the EDO has been kind to us, and I should
say Mr. Collins also, in terms of putting this team together. Art Howell is
a highly capable individual. He's leading the team from Region IV.
I was assigned as his assistant leader, and we have a very capable team here
that's distributed among both the headquarters operation and the regions.
In addition, we're going to have --
MEMBER APOSTOLAKIS: Have come you have -- well, I don't recognize anyone there
who's an expert at human performance. Shouldn't there be someone?
MR. HACKETT: You know, the team is literally so new. I have to say I believe
that Ron Lloyd has some experience in that area, and possibly Tom Koshy (phonetic),
although I could take that as an action and get back to you on it.
MEMBER APOSTOLAKIS: When we had the Athena presentations, there were usually
four or five guys sitting where you are sitting now. Maybe one of them should
be involved in this. It would help you draw some conclusions that perhaps otherwise
you wouldn't draw.
MR. HACKETT: Yeah, we have the ability to draw pretty much from what we need
on the NRC staff, you know, with the --
MEMBER APOSTOLAKIS: See, my concern is, again, that maybe we would focus on
the technical part, the hard science part, when, in fact, the failures were
not there.
MEMBER ROSEN: I think the management, George, of this lessons learned task force,
Art, Hal, and Ed, have enough experience to understand the organizational and
management factors to deal with the issues that I think you're referring to.
MR. HACKETT: I think, in fact, the focus is much more initially on the -- well,
the charter elements, what we're calling charter elements A and B on the reactor
oversight process issues and the regulatory process issues, I think, in fact,
the focus is going to be largely there.
The other three elements are important, but if I had to weight these, I think
the first two are the most important, and that's going to be the primary focus
of the task group for sure.
Anyway, we're fortunate to have this. We're also fortunate to have been given
the physically separate space on the 16th floor.
MR. GROBE: Just one other thing on the structure of the folks that are on the
committee that's important is that the committee is completely independent of
anybody in Region III or anybody in NRR that was involved in these activities.
So it's going to be a fresh look.
MR. HACKETT: In terms of how things are going to progress, I just briefly mentioned
schedule previously. We're in this preparation phase right now which really
extends to the end of June effectively. That's, you know, running from some
mundane things like getting people set up in offices to actually starting to
conduct some interviews with NRC staff and managers and, starting next week,
discussions with plant personnel at the site and also with the region.
Jack mentioned earlier there's a trip out to the site vicinity next week that
several of us will be going on. I'll mention some more about that in a minute.
The expectation from the EDO is that we're going to complete this activity in
September of this year. That's the marching orders right now. Obviously things
could be subject to change. If any new information comes to bear that would
bear on the schedule, in particular, but that's where we're heading right now.
And then I'll just end with current status. I'm sorry. This is sort of where
we are as of today. We just literally this morning completed two and a half
days worth of team orientation briefings. The team, the nine folks that I had
up there on the slide are physically here at NRC headquarters from the regions
and from the headquarters functions.
And we're all assembled in one place on the 16th floor in One White Flint.
Team orientation briefings they said are completed. We are having -- Jack is
having the 0350 panel meeting next week at the site vicinity. We are having
what Art has been calling a public entrance in the site vicinity right after
that. I believe it's late morning
MR. GROBE: It's actually before.
MR. HACKETT: Before?
MR. GROBE: Yeah.
MR. HACKETT: It's like ten o'clock in the morning, I believe.
MEMBER APOSTOLAKIS: So what is a public entrance?
MR. HACKETT: Basically it's really part of the communications plan for the task
force, is to get out to the site vicinity and let people know that we're doing
this and sort of what the expectations are going in to do that particularly
in the site vicinity.
One that I didn't put on the slide is that we are working right now on also
having a public meeting probably the week of June 17th where we'll be sort of
rolling that charter, duplicating that same kind of meeting here at headquarters
and soliciting input from anyone who's interested in providing some at that
point.
Art, in particular, has been on this longer than the rest of the team. There
have been a lot of interviews with key NRC managers who have been involved in
this, and many more are going to be in progress.
And right now the team, in fact, just this afternoon is working on detailed
review plans for the separate activities that we'll be doing. So that's where
we are at the minute.
I'm especially glad to take any inputs on the charter or any thoughts the committee
might have are welcome at any time.
CO-CHAIRMAN FORD: Thank you very much.
MR. HACKETT: Thanks.
MR. HISER: I guess what I'd like to do is take a tack that is maybe a little
bit unusual for ACRS, a little bit of a philosophical twist to things.
(Laughter.)
MR. HISER: I know you guys don't like to do that.
CO-CHAIRMAN FORD: In the course of an hour.
MR. HISER: You guys don't like to do that. Very short; three slides.
We talked quite a bit earlier this morning about use of leakage detection. I
just wanted to go over some ideas that we have. We don't have any real firm
ideas at this point, in all honesty. We're still gathering information.
We do have maybe some ideas starting to gel in terms of the philosophy of how
leak detection can fit in.
Clearly, first of all, before you determine what the appropriate inspection
methods and frequencies, what the inspection program should be, you have to
understand what it is you're trying to manage from the standpoint of where we
were almost a year ago.
When we were discussing Bulletin 2001-01, the focus was really the safety concern
of nozzle ejection. With the recent findings at Davis-Besse, that as we discussed
early this morning has really raised the bar a little bit to where leakage may
be the thing that we're really most concerned about.
And I guess the one thing that I want to impress upon the ACRS is it's not just
the nozzle base material that's of concern. Cracking has occurred in the nozzle
base material. It has occurred in the weld material. It has occurred at the
interface of the weld and the base material. It has occurred at the interface
of the butter (phonetic) and the vessel head. So pretty much all components
of this structure are at issue here.
CO-CHAIRMAN SIEBER: And none of it is allowed by the code.
MR. HISER: None of it is allowed. That's exactly correct.
But also, how can we effectively manage each of those parts is really another
key part to this. That's dependent on the state of the art, of the inspections,
and tooling and the availability of those.
CO-CHAIRMAN SIEBER: It seems to me that if you're inspecting visually for leakage,
then you've already passed the threshold in which you're in violation of the
code, and it seems to me that if you have a susceptible plant, you ought to
do volumetric and work for the 70 percent crack and fix it.
MEMBER APOSTOLAKIS: I think the leakage part is part of managing the accident
and preventing it from becoming an accident, right?
CO-CHAIRMAN SIEBER: Well, part of it is compliance with the code.
MEMBER APOSTOLAKIS: Isn't that what it is?
CO-CHAIRMAN SIEBER: That's an NRC requirement. It's a state requirement, insurance
company requirement.
MR. HISER: Well, I think it's a good lead into the next bullet.
MEMBER SHACK: Well, before you --
CO-CHAIRMAN FORD: Before you go, you skipped that one. Surely there should be,
as Jack says, there's a code that says, "Thou shalt not have a crack."
CO-CHAIRMAN SIEBER: A deep crack.
CO-CHAIRMAN FORD: Well, I meant a deep crack.
MEMBER ROSEN: Peter, we have a member of the public who wants to make a comment.
MR. LASHLEY: This is Michael Lashley, South Texas.
And I didn't bring the code book with me, but that's probably not a perfectly
accurate statement. The code allows evaluations and has certain acceptance criteria.
Cracking has acceptance criteria throughout the code. It's not precluded, and
in certain instances, specifically an example is buried pipe, it will clearly
say you can live with that as long as it's within your operational boundaries.
So it's known in the code that cracking is not a totally tabu thing. You do
have to do other measures and have other compensatory actions.
CO-CHAIRMAN SIEBER: But the reactor coolant system pressure boundary is an exception
to that.
MR. LASHLEY: Well, that's the tech spec issue. The tech spec will say.
CO-CHAIRMAN SIEBER: It's a code issue.
MEMBER BONACA: We do not wait until you have leakage in the tubes. I mean, you
go in and inspect, and you're looking at certain criteria. Now, you may have
leakage, but by the time you restart the plant you're not supposed to have any
leakage in the tube.
MEMBER SHACK: But here they don't allow operation with through wall cracks,
which is analogous to the steam generator case. I mean, you don't allow operation
with known through wall cracks.
MEMBER BONACA: But you're waiting for leakage to detect. What I mean is in the
tubes you go in, inspect, you do sampling, but you inspect and plug if your
through wall is beyond certain criteria.
MR. GROBE: I think there are two issues on the table. One is having a leak,
a through wall crack. You're clearly not permitted to operate with a through
wall crack.
But it's not uncommon to have very shallow cracks identified during IS activities
and have those be analyzed that it's safe to operate for another outage, another
cycle, and oftentimes that's exercised, and the licensee prepares for whatever
repair or replacement activities they'll do.
CO-CHAIRMAN FORD: But, Jack, surely it is up to a certain point.
MR. GROBE: That's right. That's right.
CO-CHAIRMAN FORD: You can't wait until there's a through wall crack.
MR. GROBE: Absolutely.
CO-CHAIRMAN FORD: The code doesn't allow that.
So that comes back to Jack's point. Should there not be a third sub-bullet on
the second bullet? There's a limit to the amount of cracking, non-through wall,
that you can have.
MR. HISER: Yeah, I think that's correct. The purpose of these bullets was really
to look to the point of, you know, leakage and deposits. What is allowed within
the tech specs and the ASME code, and how does this fit? How does use of visual
examinations fit within this context?
MEMBER WALLIS: Right, yes.
CO-CHAIRMAN SIEBER: Well, the way you wrote that tells me that you should look
at the code, and it tells me how far you can go, what you have to do in your
tech spec.
MR. RICCARDELLA: Peter Riccardella from Structural Integrity Associates.
You know, we're not talking about operating with known leakage here. If we find
the leakage, we fix it. We're talking about operating with some non-zero probability
of a leak, and the code doesn't prohibit that, and we do that in the primary
coolant system all over the place.
We operate with some non-zero probability of having a crack or of having a leak,
and you know, that's the issue that I think we have to have addressed. What
is the acceptable probability that we could live with, not that we operate with
leaks?
MEMBER BONACA: But you do IS in the vessel, right?
CO-CHAIRMAN SIEBER: And piping and everyplace else.
MEMBER BONACA: In piping, in volumetric inspections, and so on and so forth,
and here we're talking until now we just do visual. So with visual it means
we're waiting until we see leakage to determine that we're going to now repair
it.
CO-CHAIRMAN SIEBER: Are you going to leak? That's right.
MR. RICCARDELLA: But you know, IS of small bore piping we do visual, and you
know, we accept the fact that, for example, socket welds and small bore piping,
we have a finite probability of leakage that occurs from time to time.
CO-CHAIRMAN SIEBER: That's right.
MR. RICCARDELLA: In the primary coolant system.
MR. HISER: Yeah. I think the one context that the staff would come at this from
is the expectation previously was that these components wouldn't fail. You wouldn't
get leakage, and so maybe leakage was an appropriate method to manage for that
unlikely event.
Now, given the incidences that have been identified, you know, we need to take
another look at it. That's all we're trying to do here, is just to lay out some
of the basis for this.
CO-CHAIRMAN SIEBER: Maybe I can add additional confusion. I already wrote my
comments, and I --
(Laughter.)
CO-CHAIRMAN SIEBER: -- and I'm just waiting for you to say them.
MEMBER ROSEN: Well, Jack, do you want some more input first? We've got another
--
CO-CHAIRMAN SIEBER: Well, let me finish. I have the floor right now. Okay?
It seems to me that the susceptibility ranking curves, if they're done right,
could be a process where you decide what kind of inspection and examination
you need to do.
For example, a plant where the probability of actually having cracks is pretty
low. Maybe visual is good enough. On the other hand, if you're in the hard runner
list, you know, the most susceptible plant list, maybe volumetric is a better
deal, particularly if you can calculate, which I think that we're all trying
to do, how fast these cracks will grow, and that's basically what you do with
steam generator examinations.
You're trying to predict can I run another cycle without losing a tube, and
I think there's some value in thinking about that kind of an approach.
I would be happier if one of two things. One of them is that the database that
was used to come up with the susceptibility ranking also included information
about heats or, on the other hand, I think that whoever has a leak that appears
to come from a susceptible heat of material, write a Part 21 so that everybody
knows that here's additional susceptibility, and they can do something about
it.
So that would be my thought process as to how I would deal with these issues
you've put up here, for what it's worth, and if I get ten other people to agree
with me, we can do it right.
(Laughter.)
MR. LASHLEY: Let me make one other comment. We talked about code and we talked
about regulation. I'm going to read Criterion 14 out of the general design criteria,
which is for the reactor coolant pressure boundary.
"It is the reactor coolant pressure boundary shall be designed, fabricated,
erected, and tested so as to have an extremely low probability of abnormal leakage
of rapidly propagating failure and of gross rupture."
The code follows that same structural integrity process. It doesn't preclude
crackage or through wall leakage outright.
MEMBER APOSTOLAKIS: So don't you think though that having a through wall crack
and leakage is inconsistent with the requirement of an extremely low probability?
MR. LASHLEY: If you accepted it and just gross leakage --
CO-CHAIRMAN SIEBER: Step to the microphone and identify yourself, please.
MR. LASHLEY: Your point is well taken if you lived with it and didn't fix it
or didn't do an evaluation to show it's not a structural integrity issue.
MEMBER APOSTOLAKIS: Oh, yeah, sure. We're not talking about shooting anybody.
I mean fixing it. I think we --
CO-CHAIRMAN SIEBER: And when you talk in general terms --
MEMBER APOSTOLAKIS: Can we go to the last bullet? I'm dying to see what they
have to say.
(Laughter.)
MEMBER BONACA: You guys keep talking.
CO-CHAIRMAN SIEBER: Let me say one other thing. EDC-14 really is looking at
the reactor coolant system pressure boundary as a whole where there are some
flange gasketed joints, mechanical joints like spores (phonetic) and safety
valves and things like that, some of which leak, and so you just can't have
an absolute prohibition against leakage because some things just leak. Seals
leak; inner system leaks occur.
MEMBER BONACA: But remember those flange leaking in my judgment, they were a
measured contribution to this event because there were a fixed--
CO-CHAIRMAN SIEBER: Well, it masked the problem.
MEMBER BONACA: They masked the whole issue, and they -- so, you know, one could
even say the codes are not perfect.
CO-CHAIRMAN SIEBER: Well, I think there's a difference between leakage at some
mechanical joint and leakage because of a defective wall.
MR. HISER: Ongoing degradation does tend to cause problems.
CO-CHAIRMAN SIEBER: And go on.
MR. HISER: Right. Now, within the overall context of safety of these components
we have robust designs to minimize failures. We have quality fabrication practices
and inspections to insure that we have quality components.
The role of leak detection may be at a minimum Defense in Depth. If one had
inspection requirements that were more intensive, say, NDE, something like that,
there still may be a role for leak detection just in case something happens
different from what we expected, more rapidly than was expected. But it could
be used as a Defense in Depth approach to management.
MEMBER APOSTOLAKIS: So Defense in Depth now means that I have a redundant or
diverse barrier to something, right?
MEMBER KRESS: Not necessarily.
MEMBER APOSTOLAKIS: No? The Commission says it's the use of multiple barriers?
That's what the Commission said.
CO-CHAIRMAN SIEBER: You could use alternate techniques, too.
MEMBER BONACA: Alternate techniques or back-ups or trains, for example.
MR. HISER: Say again.
MEMBER BONACA: Redundant trains, for example, would provide you further Defense
in Depth. I mean it doesn't have to be necessarily a passive barrier. That's
only for the barrier portion
MR. GROBE: And there are three barriers. There's the fuel, primary pressure
boundary, and containment.
MEMBER APOSTOLAKIS: So anything that reduces the probability is Defense in Depth
measurable?
MEMBER BONACA: Well, I mean, it measures -- it's a broad definition.
CO-CHAIRMAN SIEBER: Sure. That's philosophical, but it sounds okay.
MEMBER APOSTOLAKIS: Well, the Commission said the use of multiple barriers,
and that's what it is.
MEMBER BONACA: No, in the protection of those barriers.
MEMBER KRESS: They didn't mention the barriers in the white table paper at all.
They said multiple -- I forget the words, but it wasn't barriers.
MEMBER APOSTOLAKIS: Measures?
MEMBER KRESS: Multiple measures to address incidents.
MEMBER APOSTOLAKIS: So this is a Defense in Depth measure against which event?
What are we talking about here? Defense in Depth against what?
MR. HISER: LOCA.
MEMBER APOSTOLAKIS: LOCA?
MR. HISER: Nozzle ejection, a redundant way of identifying the degradation that
could be ongoing.
MEMBER APOSTOLAKIS: I'm sure it is, yeah. That's what it is, yeah.
MR. HISER: Now, the industry will present their proposed inspection plan following
this.
MEMBER ROSEN: Some time after midnight.
MR. HISER: Sometime today. We started ten hours ago. So we'll push it along
here.
We did have a meeting with them about two weeks ago where they presented this
to us. Just to pull out some of the characteristics of this plan, one is it
does not consider explicitly the vessel head degradation experience at Davis-Besse.
The technical basis is still in progress. There is no report that's available
at the present time. For moderate susceptibility plants within the plan there
can be a reliance on bare metal visual examinations.
The report explicitly is limited to Alloy 600 heads with 82-182 weld metal,
and again, explicitly assumed a robust generic letter 8805 program that is effectively
implemented. And clearly, with the recent experience we've had those are some
pretty good assumptions.
I think some of the comments that the staff presented at that meeting and that
we will be transmitting to the MRP first is that the relevant visual conditions
that require follow-up examination do require better definition. Right now it
just describes relevant conditions.
Clearly, inspection methods and frequencies that they propose for the various
populations of plants requires a robust technical basis, and that's still something
that's being worked on.
The discussion of NDE, we thought that the capability and recent experience
with inspection methods and the developments that are ongoing, we thought that
should be provided somewhere in the inspection plan. The technology has improved
significantly over the last year, and hopefully that progress will continue.
As I mentioned before, right now our examinations of the J-groove welds and
some of the interfaces with the nozzle and with the vessel head are not real
detectable using the current ultrasonic methods. So that's something that requires
some work.
Another thing that isn't clear within the plan is how it's benchmarked. Clearly
we know when leakage was identified at plants. We know when circumferential
cracks have been identified, but it's not obvious that the thing is benchmarked
to when the leakage first occurred, when the first through wall cracking occurred,
but it appears to be based on discovery of the conditions as opposed to benchmarking
to the onset of the unacceptable conditions.
There have been some questions on the appropriateness of the application of
Reg. Guide 1.174 within the plan.
And finally, there is a provision in there to delay scope expansion to the next
refueling outage, and that's something we think requires significant technical
basis.
MEMBER APOSTOLAKIS: We're going to talk about this application of 1174 at some
point? I don't understand. Why is it relevant here?
MEMBER SHACK: One times ten to the minus three probable failure, conditional
probability --
MEMBER APOSTOLAKIS: Are we changing anything on the licensing basis? And we're
seeing whether it is risk significant? Is that what we're doing?
MEMBER SHACK: It says, yeah.
MEMBER APOSTOLAKIS: We're changing the licensing basis?
MEMBER SHACK: Well, no. We use that to evaluate changes in risk in a more global
sense.
MEMBER APOSTOLAKIS: Well, presumably as a result of the inspection of plant,
the change is negative.
MEMBER SHACK: Right.
MR. HACKETT: Well, no, the inspection plant admits some possibility of an increase
in risk. Otherwise you'd inspect more frequently.
MEMBER APOSTOLAKIS: Increase from what? From the previous state? See, I don't
understand the definition. Is there a change here that is permanent that is
increasing risk?
MR. MATHEWS: I would say that they're evaluating the increase in risk from this
phenomenon that we didn't know about when we originally assessed the risk from
the plant, and this is a change because now, oh, well, we could have the rod
ejection here that we didn't evaluate when we looked at the whole thing to start
with. What is the impact of that, and what is the change in risk to the public?
And what we're trying to evaluate is what is that change, and ten to the minus
six is a ballpark number that we were trying to say, you know, it would be okay
if I came in and did something to my plant and said, well, that's less than
a ten to the minus six change in the risk if I do this.
MEMBER APOSTOLAKIS: Are you doing a regulatory analysis now?
MR. MATHEWS: Me?
MEMBER APOSTOLAKIS: Whether it's worth backfitting. Is that what you're doing?
PARTICIPANTS: No.
MEMBER APOSTOLAKIS: So Regulatory Guide 1174 can be used to evaluate the impact
of previously unknown phenomena?
MEMBER BONACA: As a change, assume it is a change with respect to what was known.
MEMBER ROSEN: No, I think the question that the staff is asking is is this an
appropriate application of Reg. Guide 1.174. We haven't even heard what the
application is. The representative from the industry hasn't been given a chance
to tell us yet.
MR. HISER: And hopefully he will describe that; is that right, Mike?
MR. LASHLEY: I'll give it my best shot.
((Laughter.)
CO-CHAIRMAN FORD: Could I understand the timing of this? Obviously the industry
have come to you with a proposal. You're looking at it. What is the timing on
the resolution of these various issues?
MR. HISER: If I can get to the last slide and --
(Laughter.)
MR. HISER: -- you still have that question when I'm done, then I have failed.
We do have ongoing activities, and we have some areas of concern in general.
First of all, relative to Davis-Besse, the degradation mechanisms and rates
as described in the root cause analysis report don't have a lot of physical
evidence from Davis-Besse.
What we're looking to do is for them to back that up with work on the cavity
at Lynchburg and also hopefully some laboratory demonstrations that will give
us some confidence and reduce the uncertainty of the mechanisms and the rates
of those mechanisms.
CO-CHAIRMAN FORD: When you say "mechanisms," you don't mean mechanisms
the way I understand mechanisms. You understand the degradation process by which
things happen, but you don't know the mechanism and you can't predict it. You
don't know whether it's a generic issue or whether it's a one off issue.
MR. HISER: Right.
CO-CHAIRMAN FORD: And if it's a generic issue, when is the next one going to
be? You know it's not a major thing out there right now based on what's come
out of Bulletin 202, whatever the number is, 01, but you sure as heck don't
know what the situation would be in, say, a year's time.
MR. HISER: Right, and that's what we want to do is have the comfort of being
able to predict how things will occur.
CO-CHAIRMAN FORD: And that's what these guys are going to do urgently.
MR. HISER: Well, hopefully in order to reduce uncertainty we need these things
to occur. You know, otherwise the inspections are going to have to assume worse
case kind of conditions.
CO-CHAIRMAN FORD: Right.
MR. HISER: In order to back off of that, you know, with the necessary conservative
assumptions we need to have a greater understanding.
CO-CHAIRMAN FORD: Right.
MR. HISER: As we discussed, the industry proposal does need a sufficient technical
basis, and I think that will come over time. The staff is considering a generic
communication with Bulletin 2001-01 and 2002-01. We provided sort of a one cycle
approach to inspections, and that was sufficient. It gave us the data that we
needed to be able to go forward.
We're still not able to go forward. We're still not in a position to lay out
any long-term criteria. So this is a generic communication that will probably
be a bridge from the first two bulletins to what I would call the more permanent
requirements that would go in the ASME code or in 10 CFR, Part 50.
We are working with the staff to develop a technical basis for these longer
term inspection requirements. We don't have that ready now. I mean, that's going
to take time. I think within our action plan that's targeted for later this
year. That may be overly optimistic at this point.
And to put another idea on the table, I think that we believe that the Davis-Besse
experience has raised the bar, that the level or the type of cracking that is
I don't say acceptable, but that you really have to guard against has changed
from circumferential cracking a year ago to now even axial through wall cracking.
That's really the emphasis that we have at this point, is trying to preclude
through wall axial cracking.
CO-CHAIRMAN FORD: But to come back to my question, when are all of these issues
going to be resolved?
MR. HISER: Hopefully around the end of the year or some sort of time frame like
that is what we have worked out with the industry.
CO-CHAIRMAN FORD: This is very important. I mean if you're starting to just
do away with volumetrics and won't go through any of these kind of studied process
of when you use volumetric versus visual and you just go to visual because it's
an easy thing to do, it's major, major assumptions.
MR. HISER: I would expect that as I stated the generic communication will have
conservative assumptions. Until we have a firm understanding of things, such
as Davis-Besse, we will not take potentially non-conservative assumptions.
CO-CHAIRMAN FORD: Okay.
MR. HISER: From the standpoint of visual detection and visual inspections, I
think things will be different than what was laid out in Bulletin 2001-01 significantly.
CO-CHAIRMAN FORD: We will hear about that before it becomes a done deal?
MR. BATEMAN: I'm not sure about that. I think we're moving pretty quickly with
trying to get some generic correspondence out.
I think you can take some comfort from the fact that you're going to hear what
the industry's proposal is, but I think our proposal at this point in draft
stage is it's going to be more rigorous than what you're going to hear from
industry. I think as Allen said, I think it will be a bridge. It will probably
be more conservative than what we may ultimately end up with, but we have to
do something. We can't wait until we're through with all of this, Dr. Ford.
I mean, if we're talking about rule making, if we're talking about getting something
in the ASME code, that al takes time.
CO-CHAIRMAN FORD: But let's see what the industry have got to say.
MR. BATEMAN: Yeah, I think that's the best bet.
MEMBER APOSTOLAKIS: This is the last presentation. This must be an important
issue. Are you guys going to do this quickly?
MS. KING: We'll do this as quickly as you would like.
(Pause in proceedings.)
MS. KING: The slides for this are the las part of our original packet. And in
the interest of time, we won't be covering every individual slide that you have.
MS. WESTON: Starting at slide number 102 for the MRP part of the presentation,
yes.
MEMBER APOSTOLAKIS: One, oh, two.
MS. WESTON: The numbers are right beside ACRS 6502 and then there's a number.
MEMBER APOSTOLAKIS: Or it's four pages from the end. Go to the end and count
four pages back.
MS. KING: Okay. Peter, one thing I wanted to comment on is we have been meeting
with the staff fairly frequently, and we plan on continuing that frequency of
meeting with them on a technical level as we develop our research to get comments,
and to incorporate that in so that we don't just shop up with the final answer.
CO-CHAIRMAN FORD: Michael.
MR. LASHLEY: My name is Michael Lashley. I'm from South Texas Project.
And the first slide that we have here basically just says, yes, we met with
the NRC staff. We heard their comments, and we're actively dispositioning those
comments.
One other aspect of this just to give you real briefly where I'm coming from,
I also have the action within code space to bring these rules forward and try
to write some rules in Section 11. So myself, and I know a member of the NRC
staff, Wally Norris, is on that team. So we are trying to work together.
So we are trying to actively work it to a permanent solution.
Let me digress off of these slides real quick and show you one other slide that
maybe bridges the gap to what we were talking about, and you saw it in Pete
Riccardella's, but we have another line drawn in here that may not be obviously,
but it does speak to the Reg. Guide 174.
This slide kind of does that and also one other one. From this one, you saw
everything on this slide except this one purple curve right here. That curve
represents a one percent probability of leakage.
So you see there is a big grouping of plants in that far left-hand corner with
low head temperatures that are under one percent.
MEMBER WALLIS: One percent per year?
MR. LASHLEY: Probability of having that first leaker.
MEMBER APOSTOLAKIS: But can you explain the figure first?
MR. LASHLEY: This is the one that Pete discussed. Was that yesterday? Earlier
this afternoon, and this has on the left-hand side the cumulative effective
full power years. The red chain link that has over it the -- which color? I'm
not sure. That's kind of green. The upper one is one times ten to the minus
third, which approximately equals the 75 percent probability of leakage.
The moderate dividing line is the one times ten to the minus fourth or 20 percent
probability of leakage. So that's how we've categorized or just used that reg.
guide as a dividing line.
And then we divine an inspection program. Our attempt was to keep us under the
ten to the minus six change, to come up with an inspection program.
Now, recognize that one of the punch lines at the very end is we still have
inspection activities for this grouping in the lower left-hand corner that's
under one percent. That's at least to go after the unknown, which does speak
to defense in depth and speaks to some other issues that were brought up.
So I just wanted to show that. We'll come back to it if there's other question
because this kind of tells a lot of the story.
CO-CHAIRMAN FORD: So this essentially is you will be addressing the thing that
Jack brought up about the low susceptibility plants do visuals.
MR. LASHLEY: Yes. So we still have those elements in here. Now, at certain times
Al brought up wastage, and it's really the time line for wastage is a different
issue, and that's what wasn't explicitly addressed in our program, in our plan.
We had assumed right off the bat that generic letter 8805, it's in effect. It
is a good rule. You go read it, and it tells you exactly what to do. If you
implement it, and you all talked about this earlier; if it's implemented, there
will be no questions, but there's a desire to package this together so that
there's no ambiguity and you can see some of -- we have the ability to bring
lessons learned, bring pictures, bring training, bring a lot of things to bear
in one central document. So we're taking that feedback.
And the purpose, I mean, as we say, we assume the generic letter 8805, but we
also came up with a graduated approach for early detection, to start with low
risk, require something, require it repetitively, and then, you know, raise
the bar continuously as we move to higher and higher levels of risk.
We also believe they're very conservative for just structural integrity or the
safety issue of a rod ejection or a nozzle ejection.
This is where we start skipping a few because those have already been gone through,
but we took the technical bases. We say that the staff did not have the papers.
They were presented, and Pete presented basically the elements of it again today.
There was another technical paper that was presented by Glenn White today that's
a part of this bases, and Steve Hunt has another one.
One that we really haven't gone through is EPRI's visual guideline also, but
we bring together all of this probablistic fracture analysis, and we did sensitivity
studies to bound them to try to come up with correct inspections and correct
frequencies for the different ones to bring that to bear, and we--
MEMBER APOSTOLAKIS: Can you explain something to me? I'm missing something here.
Maybe it's me. This is a standard technical approach, you know, in an inspection
using PFM, Monte Carlo, and so on.
Then I go and I read the letter that transmits the AIT report. The first thing
they say is the boric acid corrosion control program at the site included both
cleaning and inspection requirements, but was not effectively implemented to
protect leakage and prevent a significant corrosion of the reactor vessel head
over a period of years.
And I'm sitting here trying to figure out how is this program addressing this
problem.
MR. LASHLEY: And staff brought that point up, but what you can see from that
other figure, that one again --
MEMBER APOSTOLAKIS: Yeah.
MR. LASHLEY: -- a lot of plants have done -- well, the other 68 plants have
done inspections and generally said wastage isn't an issue at my plant.
MEMBER APOSTOLAKIS: Yeah, but if there is one plant --
MR. LASHLEY: Oh, I understand.
MEMBER APOSTOLAKIS: -- where this will not be implemented, as these guys are
saying, was not effectively implemented, then the whole thing again fails. So
is this --
MS. KING: Well, there are industry activities that have been undertaken to evaluate
the implementation of generic letter 8805. We have scheduled a -- EPRI has undertaken
a conference to bring together the people that do the boric acid walk-downs
in the plant to talk about best practices, and INPO will be participating in
that conference as well.
MEMBER APOSTOLAKIS: Shouldn't that be an integral part of this inspection thing?
MS. KING: Well, as was stated in the purpose of this plan, and as the comment
we received from the staff, when we initially wrote this plan, we were depending
upon an effective implementation of the 8805 program.
As Michael stated, the words are good. It's a good rule, but we do understand
that we need to potentially -- we are working to look at the implementation
and best practices of an 8805 program.
MR. LASHLEY: And just to tag onto that, with boric acid, EPRI's guideline for
how to do this was revised just as of November 2001. So we're going to bring
all of these things back to bear at a workshop this summer, and we're going
to take the feedback we receive from the staff, and those actions are underway.
I'm the chairman of an ad hoc team under this group to try to do that, and we're
still working through that, and our time line is real tight. We would like to
bring something back through our committees by the end of next week.
MEMBER BONACA: You showed us a curve before, and you show a bunch of plants
below that purple line.
MR. LASHLEY: Right.
MEMBER BONACA: The lower purple line, and you said for those visual inspections
are justified, something of that type. What about the other plants? What are
you proposing to do for the more successful plants?
MR. LASHLEY: It is the last page of your handout. There's a flow chart, and
we're going to get there.
MEMBER BONACA: We are going to get there. Okay. So then we will just --
MR. LASHLEY: And like I said, we weren't going to go through all of our different
slides, but we'll just start doing it.
Modern susceptibility we already mentioned there was a 20 percent curve and
ten to the minus fourth or ten to the minus seventh cumulatively.
High susceptibility was using that for ten to the minus third or 75th percentile,
and that's what we meant by the Reg. Guide 174, keeping the probability under
or the change of probability under a cumulative ten to the minus sixth, which
by reg. guide standards, if you do that and a few other things, that is a risk
informed or meets the basis for a risk informed --
MEMBER APOSTOLAKIS: I have another question. My problem is what is the change.
This is a new, novel application of 1174.
MR. LASHLEY: Yes, and we're just using it to guide us. We had used probability
of leakage, and we wanted to use -- we also didn't want to be outside of, I
guess, in bad air space and risk. If I knew, you know, a rod ejection was ten
to the minus three, I should take --
MEMBER BONACA: Well, the change is similar to what has been done with 5059 for
the plants. When you discover a new condition, okay, and you want to leave with
it and you want to management it and solve it immediately, then you have to
value it under 5059 because you're changing your design basis.
MEMBER APOSTOLAKIS: But that has nothing to do with 1174.
MEMBER BONACA: Well, 1174 is in a certain way akin in that it's a risk informed
approach to the same thing.
MEMBER APOSTOLAKIS: Right.
MEMBER BONACA: You have an event. You could do things. One, you go in and just
absolutely replace the head and make a case that you have put back the plant
in the condition in which it was originally and you don't have to worry about
it for a period of time. Therefore, you don't have to do any risk evaluation.
Nothing has changed.
The other one is you want to live with it. You want to be part of this pack.
There is an increase in some risk factors there, and therefore, you are going
to justify it under 1174.
So the change is not a true change, but a change came upon you.
MEMBER ROSEN: That is the battleship in the desert phenomenon. We don't know
how the battleship got there, but now that it's there, can we live with it?
MEMBER APOSTOLAKIS: Right.
MEMBER ROSEN: And so what you do is do an analysis of what are the consequences
of that.
MEMBER KRESS: What you have is a probability of the change. If you go in and
actually find out that your probability was wrong and your detection process
showed a leak, you'd do something else.
MEMBER ROSEN: Yes.
MEMBER KRESS: You would fix that.
MEMBER ROSEN: Yes.
MEMBER KRESS: So all this is is a way to deal with the probability that you
might have approached that one time at ten to the minus sixth.
MR. LASHLEY: Right, and you'll see how once you're into the inspection program,
the results drive you then.
MEMBER KRESS: Yeah.
MR. LASHLEY: And if you're in high once, you can't get out of it. You're stuck.
MEMBER KRESS: You're there. That's right.
MR. LASHLEY: Until you replace the head.
MEMBER KRESS: Yeah, you're there. That's right.
MR. LASHLEY: So we'll go into it that way, and we did look at the J-groove weld
and put together -- because that was a concern, and it was brought up, just
the crack growth rate and things of that nature. So we looked at it from almost
the worst case to see if we needed to do something extra from what we were thinking,
and we looked at it from a worst case.
And so we used these two conditions of a circumferential crack or lack of fusion,
something that would free release that whole thing. For nozzle ejection we still
knew that it still could provide leakage or provide the environment, and those
were the comments Al said.
So we're still looking at the environment that it creates and the leakage and
the wastage. We've got to put that aside, but we did look at these two conditions
and saw that's not going to create anything worse than the circumferential crack
at the nozzle.
You'd have a circumferential crack around the J-groove itself. It physically
can't fit through the hole.
Pretty much the same thing for lack of fusion. You would have to go all the
way to still that structural margin of 300 degrees to really free release it.
So we felt we were bounded by the circumferential analysis, the deterministic
analysis that Pete's done.
So let's go into the plan. There are several slides that --
MEMBER BONACA: but you're still focusing only on the probability of rod ejection,
right?
MR. LASHLEY: That was what when we looked at --
MEMBER BONACA: I know, but now there is a new issue, which is --
MR. LASHLEY: Wastage,
MEMBER BONACA: -- one of wastage, yeah.
MR. LASHLEY: And the issue with J-groove, it will just bring it to the surface
sooner, but if a visual technique is -- we would propose a visual technique
can see it, can see the evidence, and if it's done at just the appropriate frequency,
you still won't have the wastage issue.
MEMBER BONACA: But if I remember, at Davis-Besse they had one nozzle, nozzle
number three, where they had large wastage.
There was another nozzle, number two, I believe, where there was very minute
wastage along the CNDM. Would you see that?
MR. LASHLEY: My supposition would be yes. I think you heard the 900 pounds didn't
get there overnight, and I know you saw pictures dating back further that saw
the red rust.
MEMBER BONACA: No, I understand that. I'm saying there were two areas of wastage.
One was a very large one, which may be the main source of the red. Then there
was a very thin one that I don't think a visual inspection would be visible.
MR. LASHLEY: No, we have the visual exam guideline which takes all of the other
events, the Oconees and everything. It has nice pictures and videos in there.
This is one of our reference documents to implement, to use.
And you've heard that term "popcorn." You can still have the minute,
you know, one cubic inch, the very small levels that that condition would easily
bring out. That one I think you'd still see a flow.
MEMBER KRESS: Your concern that the one leak masks the other one and --
MEMBER BONACA: Yeah, because at some point--
MEMBER KRESS: Yeah, but I don't think they would ever tolerate that kind of
leak in this system, and this is going to be so low that if you get individual
nozzles leaking, you'd know it because they're not going to have this kind of
massive leaks.
MEMBER BONACA: No, no, I understand. I'm saying in the second nozzle where there
was an incipient erosion taking place, but it was very minor. It was just very
close to the --
MEMBER KRESS: Yeah.
MEMBER BONACA: I'm just questioning --
MEMBER KRESS: Yeah, but that would be a leak that you could fix.
MEMBER BONACA: -- whether it is visible at that point. Yeah, but I'm saying
that would it be so visible.
MR. LOEHLEIN: Maybe I should comment on that.
MS. KING: Okay.
MR. LOEHLEIN: This is Steve Loehlein again.
Nozzle two does have a cavity region that maximum depth was about three-eighths
of an inch. It did extend to the surface, was visible at the surface, and through
comparison to other test data that's been available from the EPRI testing and
so forth, it's pretty conclusive that there would be significant formations
of boric acid in the region of a nozzle that looked like that, and there would
be some rust colored deposits as well because there is active corrosion products
being expelled with the boric acid at that point.
So nozzle two is actually quite far along in terms of being able to be visible
from boric acid.
MEMBER BONACA: I understand, but I'm not talking only about nozzle two. I'm
talking about another hypothetical nozzle where corrosion is going only from
the beginning of it to one third or one fifth of what we see in nozzle two.
Do you see what I'm trying to say? I mean, there is an incipient corrosion taking
place, and I'm just asking if, you know, the only criterion should be their
concern with nozzle ejection or also with incipient -- the beginning of erosion
and corrosion that would cause some coloration, but not necessarily allow the
popcorn effect.
MS. KING: I think that comment goes directly back to the work that we've undertaken,
and we're still going on, and you saw the initial presentation from Glenn White.
Our initial read on that situation is that you would have visible deposits on
top of the head prior to reaching nozzle two type wastage, and definitely prior
to reaching the cavity formation at nozzle three.
But the definite time line on that, we still have some work to do.
MR. MATHEWS: We've had 30-some nozzles that have leaked that we know of, and
every one of them has shown boric acid on top of the head, even the ones that
have had no wastage at all. And so what we're saying is that if you do start
to get wastage, you're going to start expelling stuff to the top of the head.
And it takes a period of time, and that's what we haven't quantified yet to
go from the initial leakage to getting the high flow rates that's going to generate
significant wastage.
But it's going to be visible, and if you do a visual inspection at a frequent
enough basis, you'll catch it before then.
CO-CHAIRMAN FORD: But that's not what happened at Bouget.
CO-CHAIRMAN SIEBER: Or Davis-Besse.
MR. MATHEWS: Well, I'll be honest with you, Peter. I don't know what happened
at Bouget. They weren't looking on top of their heads under their insulation.
CO-CHAIRMAN FORD: N, as I understand it at Bouget, they detected by hearing
during a -- well, they detected it during a hydrostatic test. There was no boric
acid on top of that particular --
MR. MATHEWS: Well, that's what's not totally obvious to me, that there was no
boric acid, because they had not looked under their insulation, as I understand
it.
CO-CHAIRMAN FORD: Well, I'm only reporting what was written in the paper.
MR. MATHEWS: Okay. Well, I've been trying to chase that issue down. Did they
look? And I'm not sure they did.
MS. KING: I'd like to make a further comment to what Larry was talking about
with the experience to date. We'll take the Bouget comment under consideration.
I mean, we need to get some more information.
CO-CHAIRMAN FORD: The reason why I'm bringing it up is, you know, this is a
topic that comes up, you know, in cocktail conversation time and time again,
and I keep hearing it, although we don't have any cocktails today.
Unfortunately this ugly fact destroys a beautiful hypothesis. If it really is
true --
MR. HUNT: Steve Hung from Dominion Engineering.
At Bouget, the crack, the length of the crack above the top of the J-groove
weld was two millimeters, which was less than a tenth of an inch. It was, you
know, a 13th of the length of the cracks that you had at Davis-Besse. It may
have been long enough for you to get water to create the circ. crack, the small
circ. crack that was reported, but following the model that Glenn described,
probably not large enough to create the volume of leakage necessary to create
the wastage.
MEMBER BONACA: In any event, I don't want to pursue it any further. I just want
to say that, you know, first of all, we thought that this nozzle would never
fail. Then, lo and behold, some of them began to crack.
And then we believed that they would never have circumferential cracks, and
lo and behold, we found circumferential cracks.
And then we believe that we had full control of it, and lo and behold, now we
have a hole like this up there. So I'm not an easy believer anymore. I mean,
I have to be a little skeptical about all of these promises.
MEMBER APOSTOLAKIS: This is my problem.
MS. KING: I guess I would like to comment along the lines of the experience
to date and the repairs that have been done. Typically the repair that has been
done is the Framatome what we've termed relocation of the pressure boundary,
where they go in and take out the bottom portion of that nozzle.
And in that repair process, you have the opportunity to inspect the bore, and
so far no one has identified wastage below that cut point, and you also do dye
penetrant testing to validate that you have a good place to weld.
So we do have 34 data points in the industry where we have had boric acid on
the top of the head and no known wastage behind the nozzle.
MEMBER BONACA: I understand. On the other hand, I mean, at Davis-Besse they
found the hole by pure accident because there was simply the boring and--
MEMBER APOSTOLAKIS: Well, that's my problem. I see here traditional technical
solutions to a problem that wasn't there.
MR. LASHLEY: So here's what we're going to propose --
(Laughter.)
MR. LASHLEY: -- to try to go after that.
CO-CHAIRMAN FORD: Are you going to go through these?
MR. LASHLEY: The flow chart. I'm going to go through the flow chart because
that wraps up everything on one page.
CO-CHAIRMAN FORD: Good. That's the last one.
MR. LASHLEY: It's the last page.
MS. KING: It's the last page of your handout.
CO-CHAIRMAN FORD: Now, what's in those boxes essentially is what's written down
on these low susceptibility, 100 percent reproducing --
MS. KING: Yes.
CO-CHAIRMAN FORD: The message that's in these here is important, but you're
reproducing it on this.
MR. LASHLEY: Yes.
MS. KING: Right. That is the text from the inspection plan, and this is the
--
CO-CHAIRMAN FORD: Jolly good.
MR. LASHLEY: It's probably easier to look at that.
And so you come into the plan, and let's take the low susceptibility, which
we define as less than ten effective degradation years. What we know is you
look at that big grouping on the lower right-hand side. They're all virtually
into their second tenure interval already, and we also have the rack-up of the
0201. Virtually every plant has done or is doing an inspection.
So we know we have this snapshot of that at least at baseline, and we're going
to require an additional inspection. It could be a bare metal visual or a nonvisual,
indeed, volumetric, once per ten years, and we say do that starting in your
third interval.
And our concern is, if you remember, there's such a large gap that these plants
may never cross over to moderate. If they're 560 degree head, you don't cross
over until life extension, your 52 years of operation.
So we're still requiring that group that's less than that one percent probability
of leakage to do something and to do it on a ten-year frequency moving forward.
CO-CHAIRMAN FORD: So the ten years comes from some kind of judgment. It's not
based on some criterion of some sort?
MR. LASHLEY: It had a lot of engineering judgment, and that's probably where
Al was speaking to. Like wastage in and of itself, we have evidence that we
could have found Besse six years ago, and so there is where the disconnect that
we're still working on the staff on because you don't expect the surprise down
there, but we --
CO-CHAIRMAN FORD: It's not based on some sort of analysis where you say in order
to reduce the risk by a certain amount if we need to inspect at a certain interval?
MR. LASHLEY: We did do the analysis, but if you remember Pete's curve at the
inspections, it would stay on the flat line. It would just keep bubbling up
and never come off any risk, but we knew that, but we're still going to say
you still have to do something for the unknown because we don't know what we
don't know.
But the ten-year --
MEMBER ROSEN: Modeling uncertainty or modeling completeness.
MEMBER WALLIS: The ten years is based on your assessment of what you might not
know.
MEMBER ROSEN: We require Defense in Depth. We require testing even for plants
that would not otherwise require it.
CO-CHAIRMAN SIEBER: Yeah, but this is pretty modest for a low susceptibility
plant, which is not unreasonable in my opinion. It's only ten percent.
MEMBER ROSEN: This is exactly what you were espousing, is the graded approach
to the thing.
CO-CHAIRMAN SIEBER: Yeah. I just worry about the high susceptibility plant.
MEMBER ROSEN: We'll get to that.
CO-CHAIRMAN SIEBER: The faster we get there, the happier I'll be.
(Laughter.)
MR. LASHLEY: And I'd say don't forget that at least when we first proposed that,
we still knew everybody does a boric acid walk-down every year or -- excuse
me -- every outage, and we still assumed it was more robust. So we're going
to take that. We have that comment.
Moderate susceptibility is at ten to 18 effective degradation years, and we
required once every two, not to exceed five years, and this one was an engineered
number to be less than the six that we knew about for Davis-Besse or that we
suspected.
You're going to do a bare metal visual or you're going to do a non-visual once
every four effective degradation years, not to exceed ten. And this is where
we use Pete's model and his susceptibility -- not susceptibility. I lost the
word.
PARTICIPANT: Effective inspections.
MR. LASHLEY: Yeah.
MEMBER KRESS: Some of those plants in that modern region are down near the bottom
line.
MR. LASHLEY: Correct.
MEMBER KRESS: Some of them are up near the top, the one times ten to the minus
six line.
MR. LASHLEY: Correct.
CO-CHAIRMAN SIEBER: Right.
MEMBER KRESS: Now, are you going to treat those two plants differently or they
get the same treatment? Because they're in the moderate region, both of them.
CO-CHAIRMAN SIEBER: You're profiling now.
(Laughter.)
MEMBER KRESS: I am profiling, yeah. I mean, it would make some sense to treat
those two plants differently, how close they are to that line.
MR. LASHLEY: Right. We talked about that when we received that specific comment
from the staff. I mean, there's the example of this week you're moderate. You
start back up from your outage, and by gosh, next week you're high.
MEMBER KRESS: You're across the line.
MR. LASHLEY: You're across the line. So we have evidence.
If you look at the periodicity of the exams, and most of those plants are higher
in temperature, the periodicity is two EDY versus every outage. So most of those
plants, if you are greater than 600 degrees, two EDY is only one cycle, is one
18-month cycle.
So we thought about it, and that's why we're using EDY and not years.
MEMBER KRESS: Yeah, okay. That would help.
MEMBER ROSEN: Why are you switching from EDY to EFPY? I don't understand that.
MS. KING: Well, that was to put an upper cap on those plants that accumulate
EDY very slowly, and so they couldn't go -
MR. LASHLEY: Do you want to go back to the figure?
MS. KING: Which one?
MR. LASHLEY: Heats.
MS. KING: Oh, Lord. There we go.
MR. LASHLEY: All right. Remember he had a whole series of blue lines, but EDY
goes like this. So to do one EDY it might take that long. I mean, it may take
five effective full power years if you're way over here at 560 degrees. Remember
all of these swooping -- those are EDY, the curve.
So when we used -- that's degradation years normalized to 600 degrees, but if
we keep it at, sorry, you can't go past so many effective full power years,
that was our attempt to go after the wastage.
MEMBER ROSEN: Regardless of the --
MR. LASHLEY: Regardless of it, it maxed out.
MEMBER BONACA: And, of course, it assumes susceptibilities, only temperature
dependent.
MR. LASHLEY: It's using the simplified model, yes.
MEMBER BONACA: So we're hanging a lot of things on this assumption. Again, I'm
a little bit --
MEMBER KRESS: Well, you might argue that one times ten to the minus six kind
of takes care of that uncertainty to some extent.
MEMBER BONACA: Maybe.
MEMBER KRESS: Because that's a pretty low number.
MR. LASHLEY: And you can see from this 560 degree plant to go from moderate
to high, there's still some 40-something years, effective full power years,
but that's only eight effective degradation years.
MEMBER KRESS: Now, the ones I was concerned with were these down here on the
high temperature end, say the red ones.
MR. LASHLEY: We'll get -- well, it's not on the flow chart. Let me jump in.
Any time you find a leak and it says it in the plan, you are redefined--
MEMBER KRESS: Yeah, but I'm looking at the blue ones that are that close, too.
It seems to me like some of them, a couple of those blue ones are pretty close
to that line.
MR. LASHLEY: Being blue like that means they've inspected. They're less than
probably two EDY away.
We say when you come into this plan, you're going to do this inspection. You're
going to start off doing it, hit the ground running. So that's what it was geared
for.
And we knew all of these will be moderate imminently.
MEMBER KRESS: Well, they're going to get there. I mean, that's the thing about
time being your--
MR. LASHLEY: Right, and our graduated approach is to use that effective degradation
years as the frequency, but cap it in real years so that you can't get too far
off track without coming back.
MEMBER BONACA: You have a number of red triangles there that are below the separation
between moderate and high risk. But you consider them high risk, right? Because
they already have --
MR. LASHLEY: They will be considered high risk.
MS. KING: They will be considered high risk, but also those data points are
one year old. We need to update our data points.
MEMBER BONACA: So that would go --
MS. KING: It's based on the 228 effective full power years at that count. It
would be expected to be recalculated, and I guess I'd like to point out that
those plants, well, spring '02 and the later Xes have inspections planned associated
with their Bulletin 01-01 responses.
CO-CHAIRMAN FORD: Could I suggest maybe we start to wrap up?
MR. LASHLEY: Okay. Any questions? No.
CO-CHAIRMAN FORD: I think we're al getting a bit punch drunk here.
MR. LASHLEY: High susceptibility has the bare metal visual every outage, and
it also has no matter what -- you're going to do a non-visual. You're going
to do NDE within the first four EDYs to get --
CO-CHAIRMAN SIEBER: Volumetric.
MR. LASHLEY: You're going to do it, period, and that's just to go after the
unknown.
What this also has in it if you go down below -- can you scan down? -- if you
ever find the leaker, you're going underneath. This is standard code stuff now.
You can characterize the flaw and find the extent of condition.
We do allow in this plan one cycle to complete the 100 percent look of every
nozzle under the head. So this was for that plant that found a leaker early.
You still have to go look at those, but we still felt like if you were moderate,
you still had time. If you didn't show anything above, you still had time. You
didn't have the wastage issue. You didn't have the safety issue. We could accept
a cycle before you come back in and do 100 percent volumetric of everything.
So once you're a leaker, once you're high risk, you're doing that volumetric
you're after, and you're going to do 100 percent within one cycle.
So then we would know the entire extent of condition and fix what we find. We're
using the reference flaw characteristic that directs Strosnider, and it has
virtually intersecting or circ. cracks you've got to fix, and that was the 75
percent through wall to the next inspection you have to fix.
That was short and sweet.
CO-CHAIRMAN FORD: Thank you very much, indeed. I appreciate it.
MEMBER APOSTOLAKIS: Is there a written document where all of these things are
explained?
CO-CHAIRMAN FORD: The work on the probability, French mechanics, I don't think
you were here for. The explanation for the curves --
MEMBER APOSTOLAKIS: Yeah.
CO-CHAIRMAN FORD: It's in the package though.
MEMBER APOSTOLAKIS: Is there a series of EPRI reports, or there will be?
MS. WESTON: There is a write-up on the inspection plan in your notebook, yes,
under page 117, handwritten. It's already in the notebook.
CO-CHAIRMAN SIEBER: Would this end up in Section 11?
MR. LASHLEY: Like I said, I have the action to bring it to Section 11, but we
also have a meeting this summer to try to write -- we've already presented it
twice. We hope we can bring something to start voting on this September, and
all of the intertwining, acceptance criteria and the other things that this
would go after.
CO-CHAIRMAN SIEBER: Otherwise it would have to go in tech specs in order to
make people do it.
MR. LASHLEY: Our desire in codes and fervent attempt is to get ahead of this
and do something because there's a recognized inconsistency. This is a vulnerability
that we didn't have any good inspection criteria for, none. I mean really none.
CO-CHAIRMAN SIEBER: Well, I have to think about it. This isn't really too far
away from what I was thinking anyway.
MEMBER APOSTOLAKIS: What did you say? Too far away?
CO-CHAIRMAN FORD: It wasn't too far away from what he was thinking already.
CO-CHAIRMAN SIEBER: But in order to really be efficient and practical from a
regulatory standpoint, Section 11 is the way to go, but that takes a long time.
MR. LASHLEY: I mean, we're well on a fast track. The Section 11 chairman, subcommittee
--
MS. KING: I understand what you're saying, and, you know, we have direction
from the PMMP steering committee to work fast and furious on the inspection
plan. They've reviewed it once and as Michael indicated, they would like to
see us address the staff's comments and come back out, get the technical basis
pulled together by the end of next week, and that's what we're working to do.
CO-CHAIRMAN SIEBER: Well, the staff has a couple of people on the Section 11
committee anyway.
MEMBER ROSEN: What takes time about getting the code changed is when you're
trying to get their attention with an issue, which they don't think is generic
or interesting. In this case, you don't have that problem.
MR. LASHLEY: This is a special task group that reports directly to subcommittee
Section 11. It doesn't go through working groups and such. It goes right to
the main -- to this --
CO-CHAIRMAN FORD: Bill, could I ask you? You've heard some of the concerns from
the group primarily because this is the first time you've been hit with it.
Were there any concerns that you heard raised which you are not already considering
in the list Allen put up on the board?
Do you understand the question?
MR. BATEMAN: I think I do, and I don't think there's any concerns we've heard
today form you folks that we haven't -- that aren't similar to our concerns.
Can I just briefly --
CO-CHAIRMAN FORD: Yes, please.
MR. BATEMAN: -- give a few remarks here?
I think what we accomplished today is we've given the committee an update on
the status of the two bulletins. You've got an update on the status of Davis-Besse
from the licensee, an update on the 0350 panel and the lessons learned task
force and other industry activities.
I think progress-wise since the last meeting, Davis-Besse has elected to drop
the repair options and go with the replacement head.
You had asked for data. I think industry has supplied an abundant amount of
data, and I think it's good data, a good basis for it.
I think we have from the results of the Bulletin 2002-01 inspection, which was
the bulletin with respect to the Davis-Besse head degradation, I think we have
gained assurance since we last met with you that at least at this point in time
we do not have another Davis-Besse out there. We do have some time to take the
action that I think industry has proposed here with respect to inspections and
frequencies.
I think where we're at right now is, as I said earlier, we're contemplating
some generic correspondence as a bridging document between now and when we reach
a final conclusion, and I think it will be close, but not identical to what
industry has proposed. I think it will be more conservative with respect to
frequencies and maybe inspection methods.
But I can't speak any more on that because it's in a draft form right now, and
we ave not finalized our position.
CO-CHAIRMAN FORD: Well, I would like to personally thank everybody who provided
data. I know it was a pain in the butt. It really is to do this, but if you
all recognize, the members around this table are not all experts in all subjects,
and so we ask a lot of questions, and letters that come from us are from all
of us, not just one person.
So it is invaluable that we can hear the complexity of the issues that you're
all addressing.
Normally at this point we go around the table very quickly so that each of the
members can say a couple of remarks about what they can advise and what to condense
into two hours tomorrow, and you can pass if you don't want to say anything.
PARTICIPANT: I pass.
MEMBER APOSTOLAKIS: I pass.
MEMBER RANSOM: I pass.
MEMBER KRESS: Pass.
CO-CHAIRMAN FORD: Boy, this is easy.
MEMBER KRESS: It's seven o'clock.
(Laughter.)
CO-CHAIRMAN SIEBER: Well, I guess everything that I would say now I've already
said.
MEMBER APOSTOLAKIS: Even if we make recommendations, they don't have time to
do anything about it. It's seven o'clock. They're on in the morning.
CO-CHAIRMAN SIEBER: Just tell them what to say.
MEMBER BONACA: They should go through the same material in two hours.
(Laughter.)
MEMBER APOSTOLAKIS: Maybe they should leave time for Dr. Powers.
CO-CHAIRMAN FORD: Well, let me ask the people who are presenting tomorrow, Larry
and Jim, please. Can you cope tomorrow?
MR. MATHEWS: I took our 118 slides from all of ours. I got it down to about
--
CO-CHAIRMAN FORD: Four?
MR. MATHEWS: -- 40.
MS. KING: Forty.
MR. MATHEWS: But I was intending since everybody here is going to be there tomorrow,
and Dana I think is the only one who is not here that will be there --
CO-CHAIRMAN FORD: That's correct.
MR. MATHEWS: -- tomorrow, that I was going to go through them pretty fast. If
somebody could keep their hand over Dana's mouth, we could -- you know, I'll
go through them real fast.
I know, I know.
MEMBER ROSEN: Well, you've got half of the confusion and difficulty here, and
the other half is--
CO-CHAIRMAN SIEBER: What about the rest of us, you and me?
CO-CHAIRMAN FORD: Okay. One last piece of business before we bang the gavel.
Bill has asked for a letter from the meeting tomorrow. Do I hear a suggestion
that we discuss it over dinner tonight?
MEMBER KRESS: Yeah -- no.
CO-CHAIRMAN FORD: No?
MEMBER KRESS: Not all of us are going to dinner.
CO-CHAIRMAN FORD: Okay. So is it all right if I write the draft and you can
all butcher it tomorrow?
MEMBER KRESS: Yeah.
MEMBER ROSEN: And can we discuss it the remainder of the week through Saturday
night or however long it takes?
CO-CHAIRMAN FORD: Right you are.
I thank everybody, the presenters especially. Thank you very much indeed.
(Whereupon, at 7:08 p.m., the joint subcommittee meeting was adjourned.)