ACRS Meeting on the Ad Hoc Subcommittee on Differing Professional Opinion Issues - October 12, 2000
UNITED STATES
NUCLEAR REGULATORY COMMISSION
***
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS
***
AD HOC SUBCOMMITTEE ON DIFFERING
PROFESSIONAL OPINION ISSUES
***
Thursday, October 12, 2000
U.S. NRC
11545 Rockville Pike, Room T2-B3
Rockville, Maryland
The above-entitled meeting commenced, pursuant to
notice, at 8:30 a.m.. PARTICIPANTS:
Dana Powers, Chairman, ACRS
Mario Bonaca, ACRS Member
John (Jack) Sieber, ACRS Member
Thomas Kress, ACRS Member
Ivan Catton, Consultant
James Higgins, Consultant
Ronald Ballinger, Consultant
Jack Strosnider, Division of Engineering, NRR
Jack Hayes, Probabilistic Safety and Assessment Branch, NRR. P R O C E E D I N G S
[8:30 a.m.]
DR. POWERS: The meeting will now come to order.
This is the third day of the meeting of the Ad Hoc ACRS
Subcommittee on Differing Professional Opinion Issues.
The purpose of this meeting is this subcommittee
will review the technical issues contained in the differing
professional opinion on steam generator tube integrity.
The subcommittee will be hearing from the NRC
staff today.
The meeting is being conducted in accordance with
the provisions of the Federal Advisory Committee Act. Mr.
Sam Duraiswamy is the designated Federal official for the
meeting. Ms. Undine Shoop, a staff member who is assisting
the panel, is also present.
We have received no written comments or requests
for time to make oral statements from the members of the
public.
A transcript of this meeting is being kept and it
is requested that speakers use one of the microphones,
identify themselves, and speak with sufficient clarity and
volume so they can be readily heard.
Do any members of the panel have any opening
statements they would like to make before we get started on
today's proceedings?
[No response.]
DR. POWERS: Seeing none, I will turn the floor
over to Mr. Jack Strosnider, Director, Division of
Engineering, NRR. Welcome, Jack.
MR. STROSNIDER: Thank you. Thank you, Dr.
Powers, and I want to thank the subcommittee here and
consultants for your time and effort that you're putting
into reviewing the steam generator issues. I wanted to
started off with that acknowledgment.
As Dr. Powers indicated, I'm Director of the
Division of Engineering in the Office of Nuclear Reactor
Regulation. Historically, the Division of Engineering has
had the lead responsibility for steam generator integrity
issues, specifically related to inspection and maintenance
sort of activities.
But I think as everyone is aware, steam generator
issues are really a much more multi-disciplined effort and
if you look at my second slide, which is an abbreviated
version of the agenda, as you can see, in the next few days,
we're going to be talking about a lot of different technical
issues, ranging from iodine spiking and metallurgy and NDE
to probabilistic risk assessment, thermal hydraulics, and,
ultimately, integrated decision-making.
So you'll be hearing from staff both from the
Office of Research and from NRR and from a wide variety of
different branches and disciplines.
If you look at the more detailed agenda that you
have, the names that are listed there are the principal
speakers. I would just point out that they may be relying
on other staff members to help supplement some of their
discussions, and I would remind the staff that when they're
asked to do that, to use the microphone and identify
themselves.
We've tried to arrange these issues in some sort
of logical order. Basically, these various technical issues
are all issues that feed into a more integrated assessment
and culminating tomorrow afternoon, we're going to talk
about the integrated decision-making process, which is what
I described in Reg Guide 1.174 on how to do risk-informed
license amendment reviews.
And we're going to talk about two specific
examples, one which was mentioned yesterday. That was the
Farley review, in which the NRC found the risk-informed
amendment request acceptable, extended the operating cycle.
We're also going to talk about review earlier this
year on Arkansas Unit 2 regarding a risk-informed amendment,
which the staff followed this process and we found that
amendment unacceptable. The plant subsequently shut down to
perform steam generator inspections.
The presentations that we're going to make will
cover the DPO issues. However, they go beyond that. In
response to the agenda that we were provided, we are going
to talk about additional issues.
I think that's an appropriate thing to do, because
I think it will help to provide some additional context for
the issues and really bring you up-to-date on the whole area
of steam generator regulation.
With regard to NDE and cracking phenomena, I think
there's going to be a little bit of rearrangement in the way
we present some of that. Based on the discussions yesterday
with regard to Generic Letter 95-05, I think it's very
important that we spend a little more time on that than we
might have originally planned.
Frankly, I don't think the committee was left with
the clear understanding of exactly what's in that letter or
the basis for it or our experience with its implementation.
So when Ken Karwoski talks about these issues,
he's, first, I think, going to talk about the regulatory
framework in general, but then I think he's going to talk to
Generic Letter 95-05. We want to take that as sort of a
unit before we move into other NDE and cracking phenomena
issues, because I think we need to be careful. There is a
potential to mix different issues.
And Generic Letter 95-05 is a very specific --
deals with a very specific mode of degradation and there are
some very specific requirements, and we want to make sure
that's clear for everyone.
The final thing I would mention with regard to the
agenda, I've put some hours in here which I hope roughly
correspond to what's in the more detailed agenda you have.
The main reason I did that is just to point out
that I think it's important that we do try to stick to the
time schedule. Obviously, the staff is here, it's your
meeting, and we're going to respond to what you want us to
respond to.
But as I indicated earlier, all these things, we
want to try to show how they fit together and add up to this
overall process. So I would just encourage that we do watch
the clock.
I would also point out that there has been an
extensive amount of background information provided. My
understanding is you've got about a foot-and-a-half of
paper, literally.
DR. POWERS: It's about twice that.
MR. BALLINGER: Eighty-nine pounds.
MR. STROSNIDER: Eighty-nine pounds? Okay. But
it goes to a point. There is an extensive amount of
documentation on these issues. However, I'm somewhat
sympathetic in that I suspect you might suffer from some
information overload there. So as we go through these
issues, if there are specific questions that come up, I
would offer that we could help point to the right reference
and maybe the right location of that reference that you
could study later to help address some of your concerns.
DR. POWERS: Understand the weight is not usually
an indicator of content.
MR. STROSNIDER: Okay. If I could move on to the
next viewgraph, Dr. Hopenfeld presented, yesterday morning,
a very detailed timeline that he went through. This is my
abbreviated version.
I was hoping to make perhaps a few bigger picture
observations. If I start with the lower line on here, it
talks about the DPO activities and you're familiar with when
the DPV and the DPO were submitted.
I did indicate in here one ACRS meeting that was
held in October of '97 on the DPO consideration document. I
went back and I'm not sure I've got an accurate count, but I
think I'd just mention, for the record, perhaps, that there
have been 11 ACRS meetings from 1994 to the year 2000 on
steam generator issues.
Seven of those dealt explicitly with the DPO
issues and it's several of those Dr. Hopenfeld did make
presentations.
Having said that, I recognize, in Dr. Powers'
memorandum, I think it was September 11th, to the EDO, that
the intent is to take a fresh look at these issues,
notwithstanding that there have been prior meetings. And I
think there is merit to that, also, because things have
evolved. We have information today that we'll be presenting
that wasn't available in prior meetings. So I think that's
a worthwhile thing to do.
But I did want to point out for the record that
these issues have been discussed publicly in the past.
If we could look then at the upper timeline for a
minute, I wanted to focus for just a second on what went on
here with steam generator rulemaking, Generic Letter
NEI-9706, and there's -- I guess the best way I can say it
is that there is some frustration for everybody perhaps in
how long it takes to see things happen.
But I did want to point out there are some
processes that the staff follows in these sort of
activities. The processes were established to provide
appropriate checks and balances, opportunity for public
participation, opportunities for presentations to ACRS,
CRGR, those sort of activities.
So it does take time. I continue to read and hear
about the failed steam generator rulemaking. I just wanted
to comment on that, because I guess I can understand the
perspective that, yes, the staff said they were going to
embark on generating a new rule, and that in the end, we
didn't do that.
But what I wanted to point out is that part of the
rulemaking process is to go through a regulatory impact
analysis to determine whether the rule is justified, and
there are several different ways it can be justified, but
nonetheless, we went through that.
It involved a lot of work. It involved some
groundbreaking risk assessment to support the evaluation,
which subsequently was published in NUREG-1570. It took
time to do that.
In the end, when we look at the criteria, it
didn't support the idea of a rulemaking. I don't look at
that, personally, as a failure. I look at that as we
followed the process, we looked at the results, and that was
the outcome. We also recognized when we finished that,
though, that there were some improvements that needed to be
made within the existing regulations and regulatory
framework, and, of course, that led us to the generic
letter.
There was some suggestion yesterday that
abandoning the rule had something to do with the fact that
the industry might not like a steam generator rule. That
was not why the rule didn't happen. It was because of the
reg impact analysis.
When we started looking at the generic letter, and
we did have a lot of interaction with the industry, and I
think the staff had some influence on them. I'd like to
think they saw the technical and safety merits of some of
our issues, and they developed this industry initiative
9706.
This is a good thing. I think we need to credit
the industry for taking that initiative, and I'll just give
you a few examples on that.
The existing steam generator tech specs basically
say plug at 40 percent, except for where some alternate
repair criteria would ever have been added, but for most
plants, it's shut down, do the inspection, plug at 40
percent and you can restart. There are no explicit
requirements to do a condition monitoring of the steam
generator, to understand exactly what the condition was at
the end of the cycle, or to forecast or do this operational
assessment.
By 9706, all licensees, all PWR licensees have
committed to do those and they started performing those. So
that's a good thing. Those two things alone make this a
very important initiative.
MR. HIGGINS: Jack, you said all licensees have
committed to 9706. Is that a formal commitment or informal?
MR. STROSNIDER: We have a written commitment and
I don't remember if it's on each docket or if it's through
NEI. I think it may be through NEI. But we have a
commitment that all the PWR licensees will follow the
guidelines in 9706.
I would also point out that doing that did not
relieve licensees from meeting any of the existing
regulatory requirements. In fact, we sent a letter to NEI,
which was distributed to the licensees to make that clear to
them.
So this has been a good initiative and if you
follow this timeline, it didn't put on here
direction-setting initiative 13, the Commission guidance to
interact with the industry to look for voluntary
initiatives, and it made sense, when we started looking at
what was happening in 9706 and in the generic letter, that
we should work on that approach, and that's what has been
happening.
I would indicate that -- and I'll talk a little
bit more about this in the summary tomorrow, but we have put
this effort on hold following the Indian Point 2 steam
generator tube rupture, so that we can factor lessons
learned from that event into our review of this steam
generator licensing change package, which is 9706 and more.
This actually involves some new and improved tech specs.
But we haven't gotten to approving that yet and
we're going to factor in lessons learned from the most
recent tube failure event.
So that's what we'll talk about, how things have
evolved. The other thing I wanted to point out here is a
lot of the discussions that we had yesterday and that we'll
have in the next two days deal with the risk-informed
approach to addressing steam generator tube integrity
issues.
I put a couple things on here. If you look at
NUREG-1477, there was some risk analysis in there. It
didn't deal supplementary with the area of severe accidents;
that is, core damage events leading to tube rupture, except
in a qualitative way.
I would remind people that back in the early '90s,
when we first reviewed some of these proposals by the
licensee for voltage-based approaches, that we were still
doing a very much deterministic licensing basis sort of
review. When we got to this point, we tried to factor in
some of the risk perspectives.
As you move off in time, as I indicated, one of
the major things that happened with the steam generator
rulemaking was the development of NUREG-1570 and the work
that was in there.
It was very important work, because it dealt,
again, with some sequences that weren't specifically
addressed before.
Finally, I put a milestone on here of the issuance
of Reg Guide 1.174. That's the first time that the staff
actually had Commission-approved and formal guidance on how
to apply risk-informed regulation in reviewing license
amendments.
So you have to keep in perspective that this has
been an evolving process. As such, some of the information
that you're going to hear has evolved with time. And,
again, I'd mention I think it's worth going back and
reviewing today's state-of-the-art as opposed to some of
what we can talk about historically.
So those are a couple of the major points I wanted
to make with regard to the timeline. But one other thing I
wanted to come back to, when we talk about this process and
the time it's taking to work toward this improved framework,
we're often questioned, well, what about safety.
I would point out that during that timeframe, and
I counted them up last night, the staff issued seven generic
communications on steam generators, ranging from subjects
like the importance of optimizing the inspection methods
you're using to how you deal with circumferential cracking,
to dealing with U-bend cracking, externals, degradation of
steam generator secondary side components.
So we have been dealing with the issues as they
come up, interacting with the industry through generic
communications and through our other processes for dealing
with the licensees.
Finally, I'd like to talk just briefly on, I
guess, maybe I suggestion on what I think you need to be
looking for in terms of resolution of some of the DPO
issues. You're going to hear, in some of the areas we
talked about, some specific technical answers, if I can
characterize it that way, and one example might be the first
presentation we had this morning on iodine spiking.
You're going to see where work was done,
parametric studies were performed, and we concluded that
certain assumptions are valid for performing these analyses.
But there are other issues where you're going to
hear that it's more of a process resolution and I'm not
talking about the same -- well, in some cases, it might be
the process I talked about earlier. Some of the recent
issues with regard to vibration and dynamic loads during
blow-downs.
We put in the GSI process and it's working through
and you'll hear something about that.
But there's another aspect of this. Some of the
issues that come up, particularly from a risk-informed
perspective, when we start talking about probability or
frequency of bypass events and that sort of thing, what we
concluded from some of the work was that various alternate
repair criteria for steam generator tubes or ultimate repair
methods can, in fact, influence those frequencies.
Now, we don't know, the NRC staff doesn't know
what the next alternate repair criteria is that the industry
is going to send in to us for review.
So we can't, ahead of time, come up with here's a
specific solution. But what we have said is that we will
review those things considering the risk-informed aspects.
In fact, with regard to 9706, as an example, the industry
guideline document, and the tech specs that have been
developed, the industry very much would have liked to have
had freedom to define their own alternate repair criteria,
to define their own alternate repair methods.
We have, in working on those tech specs, indicated
that, no, whenever you come up with a new alternate repair
criteria or a new repair method, you need to come to the NRC
staff for approval, and the reason we did that is so that we
can look at it from a risk perspective and determine what
the impact will be with regard to some of these
risk-informed aspects of the issue that we've been studying.
Unfortunately, we don't have guidelines out there
at this point where the industry could pick it up and do it
themselves. That's what I mean by a process resolution.
We committed to look at some of these things as
part of our reviews. I think when you hear the discussions
on Farley and Arkansas, you'll see that we're doing that.
There may be discussions about assumptions that
are made, how we do those analyses, and I don't think that
that should be a surprise, given that, again, these are some
of the first ones that we've done and those discussions are
good.
But the point I want to make is that where we said
we were going to include these things in our evaluations,
that we did that, and that's the way some of these issues
have to be addressed.
So that's basically the opening comments I wanted
to make. I'd just ask if you've got any questions for me.
If not, again, we appreciate your time, and I will
turn it over to Jack Hayes. I think he's our first speaker
on iodine spiking.
MR. HAYES: Good morning. I'm Jack Hayes, and I'm
with the Probabilistic Safety and Assessment Branch, and I
will be discussing that aspect of the differing professional
opinion which deals with iodine spiking this morning.
This morning, I'm going to be discussing the DPO
author's concern. I'm going to be discussing the staff's
assessment of that concern, but I think it's really
important to understand that it was not the DPO's concern
that had us address iodine and spiking.
That was part of an overall reassessment we were
doing with respect to accident analysis. So to understand
how we arrived at our assessment of the DPO concern, I think
it's important to understand the staff's reassessment of
iodine spiking as a whole and the conclusions which we drew
with respect to iodine spiking.
Now, the DPO author's premise is the following.
If you have a reduction in the tech spec value of primary
coolant activity level of dose equivalent iodine-131, which
is typically one microcurie per gram, to low activity
levels, that that may result in a spiking factor which is
greater than 500.
Now, if you had that situation where you reduced
activity levels and you have a spiking factor greater than
500, then the premise is that the consequences you'll have
if you have a main steam line break accident, that Part 100
dose limits will be exceeded. That's the premise.
DR. KRESS: A question.
MR. HAYES: Yes, sir.
DR. KRESS: Have these tech spec changes been
approved?
MR. HAYES: Yes. It's tech spec values lower than
one microcurie have been approved.
DR. KRESS: What levels have they been taken down
to.
MR. HAYES: I think down to probably the lowest
has been about ten-to-the-minus-two, about
one-times-ten-to-the-minus-two, that ballpark.
DR. KRESS: Okay. Thank you.
MR. HAYES: Now, in order to understand this
further, there's some background areas we believe you need
to discuss. One is iodine spiking, what is it. We think
you need to understand that how do you incorporate iodine
spiking in the calculations of releases associated for a
main steam line break, and then how does this figure in with
the voltage-based criteria that licensees implement.
Now, what is iodine spiking? Well, it's an
increase in release rate from fuel to primary coolant
resulting from a transient. Most of us like to deal with
mathematical expressions. In essence, it's a release rate
from post-trip over the release rate at steady-state or
release, if you will, pre-trip. That's a way to define it.
Now, the question is how does it occur. Well, in
order for iodine spiking to occur, you have to have a fuel
defect. If you have a fuel defect, you will get water into
the gap associated with the fuel pellet.
When that water enters, because of the large delta
T between the fuel pellets and primary coolant, some of that
water is going to vaporize. Now, at the beginning of a
transient, the fuel pellet temperature decreases. This
causes the steam which is in that fuel rod gap to condense.
Once it condenses, it causes an imbalance between
the reactor coolant and between the fuel. You get
additional water which enters. Now, because the fuel pellet
is still at a much higher temperature than the reactor
coolant, it causes some of the entering water to vaporize
and sets up a local delta P, such that as you reduce
temperature in the fuel, you're going to get water which
will enter back into reactor coolant.
That's what the spike is. Now, as the fuel
temperature decreases, eventually you will shut off the
spike.
Spikes occur, power transients, startup, shutdown,
typical occurrences. These are included in the analysis
which we perform for steam generator tube rupture analysis
and main steam line break accidents.
DR. CATTON: This kind of says that the post-trip
release is somewhat independent of the steady-state release
rate, because you're really sort of clearing out the iodine
from the fuel itself.
MR. HAYES: Yes.
DR. CATTON: It's kind of separate.
MR. HAYES: Yes.
DR. CATTON: That would be why the ratio would
increase when you reduced.
MR. HAYES: That's correct.
DR. CATTON: Do you account for this?
MR. HAYES: Yes, we do, and we'll be going right
through it right now. It's a good lead into how we do these
calculations.
When we do a main steam line break accident or a
steam generator tube accident, we usually presume that the
calculations are done at tech spec values. For example,
normal operating primary to secondary leak rate, this value
is typically one gallon per minute.
The primary coolant activity level is also a tech
spec value. This is typically one microcurie per gram of
dose equivalent iodine-131. And then the spiking factor
which we assume in these calculations is a factor of 500.
Now, for this particular scenario, the dose acceptance
criteria is 30 rem thyroid for the exclusionary boundary,
EAB, the low population zone, LPZ, and for the control room.
Now, one thing I would like to point out is that
when we do these calculations, typically we don't approach
this value of 30. In other words, in essence, most of the
PWRs, they're probably anywhere from a factor of three to
ten below these values. That's where you typically are in
terms of the evaluations in normal situations such as this.
Now, in order to perform these calculations, it's
necessary to determine the equilibrium release rate
associated with the fuel. For example, it might be that
four microcuries per gram of dose equivalent iodine-131,
that the release of iodine-131 is five curies per hour.
When you factor that into a main steam line break
or steam generator tube rupture accident analysis, since the
spiking factor is a factor of 500 larger, this goes into the
release rate.
So in other words, when you start this accident,
you have your primary coolant activity level and then you
start releasing from the fuel into primary coolant. In this
case, it would be 2,500 curies per hour of iodine-131.
DR. KRESS: A question. To get that five curies
per hour out of the .4, which is what's measured, you have
to favor in the cleanup system.
MR. HAYES: Yes.
DR. KRESS: And the decay constant for iodine.
MR. HAYES: Right. The terms, if you will, you
have a release into the fuel, but you have a removal, and
the removal consists of decay, let-down flow, and also
primary to secondary leakage.
So all three of those factors are utilized to
arrive at the five curies per hour.
DR. KRESS: The major one being the cleanup
system.
MR. HAYES: Yes. And depending upon the isotope,
also, decay. Primary to secondary leakage --
DR. KRESS: You deal with more than just I-131.
MR. HAYES: Yes.
DR. KRESS: You use the other isotopes.
MR. HAYES: Yes. All five isotopes of iodine are
utilized.
DR. KRESS: All five isotopes.
DR. POWERS: A couple of questions. On the
previous slide, you indicated most calculations were done
with the tech spec limit, which you cited as one microcurie
per gram.
MR. HAYES: Yes.
DR. POWERS: And now you've switched to .4
microcuries per gram. Was there a reason for that?
MR. HAYES: The only reason I did that is because
in doing the slides and the preparations, one of the
examples I had was that .4. Let me give you what an actual
number is. I've looked up an actual number. I'm just doing
an amendment associated with Watts-Barr. And this number,
one microcurie is, for Watts-Barr, corresponds to like 13.8
curies per hour release rate and the 500 times that value is
like around 6,900 curies per hour.
So that was just chosen. I probably should have
put an example of one, but that's what the number is. Also,
if you like numbers, the typical primary coolant activity
level at one microcurie for Watts-Barr, that's 161 curies of
iodine-131 starting out. So that gives you some numbers.
DR. KRESS: Is there a wide variation in the tech
specs?
MR. HAYES: No.
DR. KRESS: It's generally around one.
MR. HAYES: Yes. In essence, for all plants which
do not have the alternate repair criteria, the value is, in
essence, one.
DR. POWERS: There is a peculiar habit of iodine,
it usually gets called hideout. Did you attempt to account
for hideout?
MR. HAYES: No. We don't have any special
function that includes hideout.
DR. POWERS: What does DE stand for?
MR. HAYES: Dose equivalent. I'm sorry.
MR. SIEBER: Could you tell us again what the
basis for the 500 spiking factor is and is it the same for
every plant under every condition, with any amount or no
fuel leaks?
MR. HAYES: Right. Okay. The value of 500 is the
same for all plants which was utilized. I cannot tell you
the basis, other than to say it's in the standard review
plan. I can't tell you how it was arrived at back probably
in the '70s.
I presume that someone did an enveloping estimate,
but I don't know what the basis for it is.
MR. SIEBER: Would that not be the most important
factor to consider in trying to figure out what the dose
equivalent at the exclusionary boundary and the control room
would be? It would seem to me to be the most important
thing.
MR. HAYES: No, it isn't. The most important
thing is your primary coolant activity level and also the
primary to secondary leak rate, because -- let me go back to
a slide.
MR. SIEBER: The primary coolant activity level,
the spike actually is where the dose comes from and if it's
500 times greater than what you would ordinarily have as a
dose commitment from primary coolant, that's a substantial
increase.
DR. POWERS: We're talking about a release rate
and not -- he's not multiplying his primary coolant
concentration by 500. He's multiplying his release rate by
500.
MR. HAYES: The other thing I think is important,
this is the definition of a spiking factor. We have two
terms. We have one as the numerator and one as the
denominator. For example, if the denominator is extremely
small, this spiking factor will be very large.
So the absolute value of the spiking factor isn't
as important as is really the primary coolant activity that
you have and the primary to secondary leak rate.
In further discussions, when we go into the
parametric analysis that we did, we hope that we will be
able to show you that that is indeed the case.
MR. BALLINGER: A question. You say that you
don't really consider hideout, as it were.
MR. HAYES: Yes.
MR. BALLINGER: But do you look at the sort of
steady-state, if you want to call it steady-state, iodine
concentration in the system as a function of time versus how
the plants operate, to try to get an idea of whether you're
using an average which has a lot of uncertainty in it or
not?
I mean, how do you arrive at the .4, for example?
MR. HAYES: The value of one is a tech spec value
and plants do not operate even close to that. The reality
of the situation is this. When plants start to get at
around ten-to-the-minus-two, they get real, real antsy and
they start to take action.
And, see, the presumption that we have with this
particular analysis is that you are already at one
microcurie.
MR. BALLINGER: But the proposal is to reduce that
in some cases.
MR. HAYES: Yes.
MR. BALLINGER: So a value that's supplied to you,
that comes in at some number less than one.
MR. HAYES: Right.
MR. BALLINGER: But how do you evaluate whether or
not that's a number which you believe?
MR. HAYES: You mean whether they're going to get
to that number?
MR. BALLINGER: No. Whether the -- oh, even the
lower one is a tech spec number?
MR. HAYES: Yes.
MR. BALLINGER: Okay.
DR. KRESS: The assumption seems to be that if you
have a number in the tech specs that is an allowed number,
that there is a potential then for when the accident occurs,
the steam generator tube rupture, that you may be at that
tech spec number, since it's allowed.
MR. HAYES: That's an assumption made in the
calculation.
DR. KRESS: So when you do the design basis
accident calculation, that's why you use the one or whatever
the tech spec number is and not the actual, because the
actual is not really of interest.
MR. HAYES: That's correct. That's very good.
Voltage-based criteria. As Jack Strosnider --
DR. CATTON: Excuse me. I kind of got lost in a
little bit of this. You use the number one when you do your
calculation.
DR. KRESS: Or whatever is in the tech spec.
MR. HAYES: Or whatever the tech spec value is.
DR. CATTON: Right. So if somebody comes in and
says, gee, I want you to relax a little bit, I'm going to
reduce my tech spec value to .1 or .5, in reality, nothing
has changed because they're already operating at a much
lower number.
DR. KRESS: Something would have changed, because
if they --
DR. CATTON: If they operated at one, it would
change, but if they operated at .1, then nothing really
changes, except the fact that they've reduced the --
DR. KRESS: There is a virtual change, and that is
that if they did approach this new number, they would have
to shut down and do something.
DR. CATTON: We heard that they don't.
DR. KRESS: I know, but that's -- but it's a
virtual change. They would have to if they did.
DR. POWERS: A virtual change is no change at all.
DR. KRESS: That is a change.
MR. HAYES: I understand some of your quandary.
Let me share some experience with respect to generation of
the steam generator rule.
The value of one is not a value that is frequently
met. People maintain their concentrations much lower than
that. So, if you will, it's not in a usual operating area.
In the discussions with some of the utilities, as
part of the steam generator rule, they had mentioned to the
staff, they said, hey, we think your evaluations are too
conservative, and what the staff -- we went back and
reassessed that.
Well, what we found is when you reassessed it,
hey, one of the things you licensees and we can do is let's
change the tech spec value, and the operators said, no, we
don't want to change the values. We want that margin.
So what you actually find out is plants don't want
to change that value of one unless they have to and what
happens is with the ARC amendments, in essence, they're
forced to change it, but they don't want to do that. They
want to keep that value of one.
As another example, there's another tech spec
value which is a maximum instantaneous value, that we're not
talking about today, which is a value of 60. That value, no
one has come close to it.
We've had three values above 15 in the whole
operating time period. The highest value is 18. They don't
want to change the value of 60. The value of 18 was in
1972, in Ginna.
DR. KRESS: In essence, then, what I read into
that statement is that by changing the tech spec value, you
have eroded the margins. You have to have. The question
that comes to mind when you make that statement is how much
margin do you need in design basis space.
MR. HAYES: You've eroded the operating margin.
Dose margin, you're still at the same number.
DR. KRESS: Well, you've eroded the margin to how
well you've protected against receiving a particular dose at
the site boundary or the control room, which is what you're
interested in, how well you control the potential for having
that dose.
So you've definitely eroded that margin when you
lower the tech spec value, there's no doubt, in my mind,
because in essence, you allow more leakage to meet the
values.
DR. CATTON: You're swapping real safety for
virtual safety.
DR. KRESS: Yes, exactly.
DR. POWERS: You can do that on virtual changes,
you can do that on margins, too. He has enormous capacity
for --
DR. KRESS: So this all gets embroiled in how much
margin do you need in design basis space and why did you
have the margin you had in the first place and those types
of questions.
DR. BONACA: It seems the issue is with the
leakage rate, right?
MR. HAYES: That's correct. That's correct.
DR. BONACA: The accident analysis is used on GPM
and that's why the number is in tech specs, but some plants
operate with less than that.
MR. HAYES: That's right.
DR. BONACA: And if they have problems, in fact,
meeting those limits, especially control room issues, leads
you at times to the need to tighten up that leakage rate,
which means you have to perform your analysis with a lower
leakage rate.
But I'm trying to understand how that affects the
margin issue and the tradeoff.
MR. HAYES: And I think you all are doing very
good, well, leading right into the succeeding slides,
because in this voltage-based criteria, that's the tradeoffs
we're getting into.
As Jack Strosnider mentioned, he said that the
tech specs, as they're written, at 40 percent through wall,
you have to start plugging tubes.
What the voltage-based criteria does, it allows
tubes that you would normally be plugging to remain in
service, but there's a tradeoff with that. That is, it's
postulated that if you have a high pressure transient, such
as a main steam line break accident, you're going to cause,
because of the high differential pressure across those
tubes, you're going to cause those tubes to open up and
they're going to leak at a certain level.
For example, let's say that you have a main steam
line break associated with a given steam generator and let's
say in that steam generator, you have 100 tubes which have
this voltage-based criteria.
If it is postulated that each of those tubes which
has that criteria would leak at half a gallon per minute
when exposed to that pressure, then for those 100 tubes, you
would have a 50-gallon per minute leak.
Now, that's a new source. We hadn't had that
before. The source we had was normal operating primary to
secondary leakage, but now we've got what we refer to as the
accident-induced leakage, and that value is like, for
example, in the case I used, 50 GPM. So that has to be
added into your source.
DR. KRESS: If I view this from the perspective of
Reg Guide 1.174, this is sort of a change to the licensing
basis that increases risk a little, it's a question of how
much. The Reg Guide 1.174 calls for maintaining the
defense-in-depth philosophy and it also calls for preserving
margins.
I'm not sure I quoted it correctly with those and
that's why I'm looking behind me. It seems to me like this
increase in risk, it's bound to reduce your defense-in-depth
a little bit. I don't know how much. I don't know how
that, quote, preserves the defense-in-depth philosophy and
it also erodes the margins, so I don't see how it preserves
the margins, which are sort of part of the whole integrated
analysis.
Maybe someone back here could speak to those
questions.
MR. HOLAHAN: This is Gary Holahan, from the
staff. I think in Reg Guide 1.174, the principals are laid
out to keep any risk changes small and to preserve
defense-in-depth philosophy and to preserve sufficient
margin, but the discussion of defense-in-depth and margin
wasn't meant and doesn't mean that we're not prepared to
make any changes.
So there's a judgment involved and there's some
guidelines involved as to how much margin ought to be
preserved and what does it mean to be preserving
defense-in-depth.
I think what you really need to do is to look not
only at this design basis case, but look at the implications
of any of these changes from a severe accident risk point of
view.
I think Jack isn't going to cover all of it, but
when Steve Long speaks later, you'll see the whole picture
of how we consider these issues and I think by the time we
get to tomorrow's examples that Jack mentioned this morning,
the Farley and the Arkansas amendments, you'll see, in fact,
that there is an explicit discussion by the licensee and by
the staff on safety margins and defense-in-depth and risk
implications of these sort of changes.
DR. KRESS: Thank you.
MR. STROSNIDER: I'd like to provide one other
perspective, too, which is with regard to the type of
degradation that's being dealt with here, which is outside
-- stress corrosion cracking at tube support plates, again,
a very specific form of degradation.
There was some discussion yesterday and there will
be some additional discussion today about the difficulty in
using eddy current methods to size racks and that sort of
thing.
Before the voltage-based criteria went in, what
licensees were doing was they were attempting to size these
indications and they were leaving them in service. So when
you look at the delta between what the practice was and
what's coming here, you need to ask yourself the question
which of these is really providing a more reliable approach.
Now, I will acknowledge there is another approach,
which would be plug every tube that has an indication, which
is what they typically do with stress corrosion cracking.
But if sufficient margins can be demonstrated by
the industry, then that's not necessarily the way you have
to go. So you need to look at this in terms of the
perspective, you know, do you have greater uncertainty
trying to size these cracks in terms of their actual
physical dimensions, which is what people were doing and
there was large uncertainty in that, and, to a certain
extent, dealing with accident-induced leakage was, to some
extent, probably just acknowledging reality and specifically
dealing with it.
So that's a different perspective. Steve, have
you got something you wanted to add?
MR. LONG: This is Steve Long, with the staff.
Just a couple of things to make sure we're clear on. When
Jack does this calculation in design basis type analysis,
he's assuming that leak rates that are measured as if
they're in the free span are going to occur.
So he essentially has data from tubes that they
tried to figure out what the leakage would be if the area
that's normally captured by the tube support plates is
exposed in the free span and then given the delta P change
from the steam line break.
This is not something you just throw into a risk
assessment in that manner. So you need to think of this
somewhat differently than the way you would do a
risk-informed application. These are not risk-informed
applications under Reg Guide 1.174 and the leakage that he's
using assumes that every flaw that's under a support plate
is instead in the free span.
So it's sort of a virtual leakage calculation.
MR. STROSNIDER: A conservative one.
MR. BALLINGER: But the vendors, their correlation
assumes no restraint by the support plate anyway.
MR. LONG: That's correct. But you would have to
have what fraction would be --
MR. HAYES: The question becomes, well, what is
the impact of this voltage-based criteria. Well, the goal
is to minimize the number of plugged tubes.
DR. POWERS: Whose goal is that?
MR. HAYES: That's the licensees' goal. That's
the licensees' goal. And because what happens, if you
continually plug tubes, it's obvious that you have to
de-rate your unit, and that's what they don't want.
Now, as we mentioned at the start, we had two
criteria. You had the one gallon per minute primary to
secondary leak rate and you had the one microcurie per gram
of dose equivalent iodine-131. Those are an equilibrium
now.
If, all of a sudden, you've raised that primary to
secondary leak rate, then you have to lower the allowable
primary coolant activity level if you're going to have this
larger leak rate and still maintain your doses.
So that's what they do and that's what they're
trying to achieve.
DR. POWERS: Is there a hazard, safety hazard
associated with plugging tubes?
MR. HAYES: If you plug tubes, you remove the
capability of the steam generator to remove heat from the
core and you reach a point where you have too many tubes
plugged, you have to de-rate your unit. Now, if the
question is deeper than that, I'm going to have to refer to
someone may be from the Division of Engineering to answer
that question.
MR. HOLAHAN: Let me try. It seems to me that
when tubes get plugged, the plant meets the same
deterministic safety criteria in terms of mechanical
requirements, in terms of thermal hydraulic analysis, in
cases, they have to redo the LOCA analysis. In effect, they
meet all the same requirements.
I'm not aware of any risk assessments that
indicate that taking tubes out of service by plugging them
introduce any risk changes.
DR. POWERS: No accident initiators associated with plugging
tubes?
DR. BONACA: The only sensitivity I think I could
think of would be if you had very uneven plugging in
different steam generators. That, you would have to --
those kinds of issues are considered, have to be considered.
DR. POWERS: So if I plug all the tubes on one
quadrant of the steam generator, I get some sort of problem.
MR. SIEBER: You'd get an offset in the core,
because --
DR. BONACA: You would get --
MR. BALLINGER: Between one steam generator in one
loop and another steam generator in another loop where you
have a very odd -- a very large difference in number of
tubes plugged, then you alter that.
MR. HOLAHAN: Those are the kinds of issues that
are dealt with in the analysis.
DR. CATTON: So, in essence, what you're doing is
you're just making sure the iodine that's sitting around
ready to get out is the same.
MR. HAYES: Right.
DR. CATTON: You reduce the level in the primary
system. You allow a little bit more to be dumped in. The
more that's dumped in relates to the leakers. The leakers
is tied to voltage.
DR. HOPENFELD: Right.
DR. CATTON: It sounds like a rather simplistic
calculation. Do you put any uncertainty on the steps in
this in order to --
DR. HOPENFELD: No. It's done in a deterministic
manner. There's no uncertainties put on it. Now, realize
that the criteria here is 30 rem thyroid, which is ten
percent of Part 100 limits.
DR. POWERS: That's the final.
MR. HAYES: That's the final, yes.
DR. POWERS: You're playing around with some game
in the middle somewhere.
MR. HAYES: The only uncertainty to that, I think,
is utilized with respect to leakage associated with the
tubes.
DR. POWERS: There is no consideration in the
analysis that the events that lead or are associated with
either a steam generator tube rupture or main steam line
break could, in fact, cause a rupture to the cladding on
some --
MR. HAYES: That's correct. There is no
consideration for that.
DR. CATTON: When you do that --
MR. HOLAHAN: Excuse me. I know why. It's
because there is a requirement that for those events, the
fuel continues to meet specified acceptable fuel design
limits, which are intended to not induce additional fuel
failures.
Of course, the fuel failures that preexisted are
covered in the analysis and the spiking. But you would
expect no additional fuel failures because fuel is analyzed
under those thermal hydraulic conditions to meet its design
requirements.
MR. BALLINGER: Is there any data on spiking for
actual delta P's which would exist during the main steam
line break?
MR. HAYES: No.
MR. BALLINGER: As opposed to operational delta
P's, which is all the stuff that I've been reading.
MR. HAYES: No, there is not.
MR. BALLINGER: So what is your judgment with
respect to the spiking factor when you make the jump between
what's been measured and what actually might occur?
MR. HAYES: We're going to discuss that and if we
don't answer your question when we get to that point,
please, bring it again.
DR. CATTON: So under this new approach, how do
you calculate the spiking factor? Is it this 500 times the
new lower primary system value?
MR. HAYES: That's correct. That's correct.
DR. POWERS: I guess I don't understand that.
DR. CATTON: So it goes way down.
MR. HAYES: That's correct.
DR. CATTON: And what happens to what leaked into
the primary system? I don't quite follow -- I'm having a
little bit of a problem with iodine conservation.
MR. HAYES: The iodine is not at the one anymore.
It is at some lower value.
DR. CATTON: Reduce the level.
MR. HAYES: Reduce the level.
DR. CATTON: Now I dump some in from the fuel
because I've got leakers.
MR. HAYES: That's correct.
DR. CATTON: I now multiply that new value by 500.
MR. HAYES: That's correct.
DR. CATTON: Or do I multiply the 500 times the
steady-state value before it leaked?
MR. HAYES: Before it leaked, because, for
example, I think the question that you're getting to is
saying, hey, can I have this massive amount of iodine into
the gap that's just ready in there to break loose and it
hasn't, because it hasn't been exposed to this differential
pressure or this transient, whatever it is.
And the question is, if you had that activity
available for release, it would already be showing up into
the primary coolant, if it's through the defects.
DR. CATTON: What it really gets down to is you're
reducing -- if it's 500 times a lower number, where does the
other go?
DR. BONACA: If I understand it, it reflects the
conditions of the core. If you have a very clean core where
you have no leakage, because assume, for example, you have a
brand new core coming in, where every defect has been
replaced with a new fuel rod, so there are no defects.
DR. CATTON: So you wouldn't have much iodine.
DR. BONACA: You would have not much iodine. Then
if you have the accident, unless you postulate it, the
accident causes the cracks in the cladding, which we don't
believe that is the case.
DR. KRESS: It commits you to better fuel,
basically.
DR. POWERS: I don't think so. It doesn't seem to
me it does that at all; that there are multiple ways that I
can get my primary coolant concentration down. One of them
is I can buy a better cleanup system.
DR. KRESS: Yes, but he said you account for that.
DR. POWERS: He accounts for that.
DR. KRESS: You go to the rate and you account for
that.
DR. POWERS: I'm accounting -- I want to get my
steady-state concentration, normal operation.
DR. KRESS: You increase the cleanup rate. That
doesn't help you any. It doesn't help you any in here at
all if you do it that way, because --
DR. POWERS: That's right, that's my point is it
doesn't help you at all.
DR. KRESS: No, but they account for that, he
said.
MR. HAYES: We have factored that removal rate
into the production rate.
DR. KRESS: Yes. They work on production.
DR. POWERS: I understand this all. My point is
that it does not commit you to better fuel. There's another
way around the barn here. So you don't end up multiplying
--
DR. KRESS: You're basically right, but you still
have to meet the dose limits.
MR. BALLINGER: But by having better fuel and by
having, for a given cleanup system, a lower iodine
concentration, you have a lower release rate, as well.
MR. HAYES: That's correct. And I think it's
important to realize that you cannot commit to these lower
levels, like ten-to-the-minus-two or so, because as I
mentioned, they're not comfortable at those levels, unless
you have good fuel.
Because what you see typical plants operating at
now is like ten-to-the-minus-three, ten-to-the-minus-four.
That's typically what they're operating at. And if they
start to approach ten-to-the-minus-two, they get antsy.
MR. BALLINGER: It's about ten-to-the-minus-four,
roughly, per failed fuel element. So it takes about one
failed fuel element for them to exceed their -- to make them
get antsy.
MR. SIEBER: One rod.
MR. BALLINGER: One rod, yes.
DR. BONACA: I think, historically, we have to
look at -- I mean, plants used to run with several defects
years ago. Today, I mean, since then, there has been the
goal of INPO of zero defects, and I think the utilities have
been very committed to it and it's very unlikely that you
find plants --
I mean, today, if you find a situation like a
Seabrook, where they had eight fuel failures, they had
measure inspection and evaluation of why you get those kind
of issues, and repair it. So I'm saying that reflects, in
part, the way that the fuel is being treated today.
MR. HAYES: I think if you look at the spiking
data, you can even see that. You can see that the levels at
which the spikes have occurred and the values of the spike
have really changed with time. You just don't see them
occurring.
DR. KRESS: Could you address the 30 rems as
opposed to the 300? Was that just chosen for margin or was
there another reason?
MR. HAYES: You know, this is my own
interpretation. There is no basis of fact. My
interpretation is there is a calculation which is done at
the 60 and that has the full Part 100 value. As I mentioned
earlier, no one has ever come close to the 60.
I believe that the reason that you have the value
of 30 is to account for uncertainty both with the spiking
factor and the fact that people have been at one and higher
than one.
So I think that was probably what was done,
because at that time, in terms of accident analysis, a lot
of thought was given to the more that you have the potential
for a release, the lower you put the limit. For example,
that's why a fuel handling accident has a lower one.
MR. HOLAHAN: I think there's also another factor,
because the approach of using some fraction of Part 100 is
used in some other cases, as well, and I think Part 100, the
guideline is set up for really a maximum hypothetical
accident, which is considered to be extremely unlikely.
And recognizing that some events, like tube
ruptures are much more likely --
DR. KRESS: They're much more likely to happen, so
you factor that in.
MR. HOLAHAN: You'd factor that in.
DR. KRESS: Makes sense.
MR. HAYES: It's important to understand that the
staff's reassessment of iodine spiking was just one part of
a total reassessment in terms of the way we did main steam
line break accidents and steam generator tube ruptures as
part of the rule, steam generator rule initiative, because
we had a situation where industry is saying we're too
conservative.
We have the DPO, which says we're not conservative
enough. So we had the great opportunity to make no one
happy. So what we wanted to do, you know, we were really
truly interested in the 1996-1998 timeframe, really
reassessing how we do the accident evaluation.
So one part of this was with respect to iodine
spiking. Now, the industry was proposing some iodine
spiking models. One was in a report by Postma, which was an
empirical model. Another was a first principle model by
Lewis and Iglesias, which the staff reviewed, but the staff
determined was insufficiently mature, didn't have adequate
V&V, and did not predict a priori what the spike would be.
So then we went to look at a couple articles by
Adams, articles he did with Sattison and Atwood, and the
data that Adams generated was really also included in both
the Postma and Lewis and Iglesias reports.
Now, the Adams and Sattison article looked at 58
events. The spiking factors ranged approximately two to
slightly over 900. Issue activity levels ranged from
roughly ten-to-the-minus-three to about one microcurie per
gram. The spiking factors in three cases were greater than
500. Two of those were in the range of about 900. The
activity levels associated with those spikes were roughly
two-times-ten-to-the-minus-two or less, and the maximum
activity in any case was 3.5 microcuries per gram.
DR. POWERS: I'm going to have to ask you what you
mean by maximum activity.
MR. HAYES: That the activity, at the end of the
spike, the maximum activity which was obtained in reactor
coolant, measured in reactor coolant, was 3.5.
These were Adams' and Sattison's conclusions.
They concluded that, first of all, large spiking factors
tend to be associated with small coolant activity levels and
small iodine release rates and that to assume that you have
a spiking factor of 500 in association with a dose
equivalent iodine value of one microcurie per gram is overly
conservative.
They recommended to expand the database to
determine and reduce uncertainties.
DR. POWERS: It seems to me the codicil that the
spiking factor of 500 in association with one microcurie per
gram is very important. That's a non-negligible statement
there. That doesn't seem to be in my viewgraph.
MR. HAYES: The viewgraph, I think it stopped --
this --
DR. POWERS: That's the one I wouldn't leave out.
MR. HAYES: Right. Well, in proofreading --
you're correct. In proofreading, that thought was missing
and that was a very important thought.
DR. KRESS: That doesn't address the question of
how conservative, if at all, the 500 is at some other --
MR. HAYES: That's correct. At this stage, it
does not address that.
DR. KRESS: Okay.
MR. HAYES: And Adams --
DR. KRESS: We always thought it was conservative
at one microcurie per gram. I don't think that was ever a
question.
MR. HAYES: And you're correct. The issue which
Dr. Hopenfeld has raised is a new issue and wasn't really
considered at the time that Adams was doing this work, per
se.
DR. KRESS: I see. Okay. That helps.
MR. HAYES: Because I believe, and Dr. Hopenfeld
can correct this, but I believe that this is work that Adams
was doing for you in the Office of Research at the time.
DR. HOPENFELD: He was doing this for me and I
just thought there was something missing there. Yes. He
was doing this thing for me and we went through all the data
that was the conclusion of, and I'm glad you caught that one
microcurie, because that was really the key thing.
But we didn't carry that beyond that point, for
one reason or another, and that's what -- but I thought, at
the time, they should have been carried, but we didn't. So
we stopped right here.
MR. HAYES: The next report by Adams was --
DR. HOPENFELD: Excuse me just one minute, if I
may. You left the impression that I'm trying to be more
conservative, that I'm saying that the utilities are not
conservative enough. That's not really my point. I don't
know, they may be still very conservative.
My point is that what you are doing is not
justified, just arbitrarily taking that 500 and leaving it
there while you're doing something else. That's my point,
my main point. There's no technical justification to it.
I don't know whether you're conservative or not.
MR. HAYES: The next report by Adams and Atwood
looked at spiking and the information associated with LERs.
What they decided to do is they decided that they would
bound each spiking event. They decided that the -- they
postulated that the maximum dose equivalent iodine-131,
which is typically measured between two and six hours after
the event, could be no greater than a factor of three higher
than the measured value.
So they decided that they would bound those
values. Let me show you --
DR. KRESS: Where did the factor of three come
from? Because that certainly is not decay.
MR. HAYES: I believe that the value, that they
came up with a value of three was probably associated with
the pressure differential associated with the vent.
So they said, hey, we're going to presume that
there is a linear relationship between the pressure
associated with that particular --
DR. POWERS: Didn't they raise questions about
exactly when the sampling was done in the solution and there
may have been some decay?
MR. HAYES: There were three graphs that they
presented, one of which did not include -- well, excuse me
-- included the factors of three, but also corrected
presumed -- I think that it was six hours and moved it back
to two hours.
MR. BALLINGER: I don't think it says anywhere in
what I've read that that factor of three is related to an
increase in delta P. That's why I asked the earlier
question.
DR. POWERS: I think it's a sample in their
analysis, a question that they had about the data that were
available to them. They came up and said, well, it can't be
any worse than a factor of three.
DR. KRESS: It's uncertainty in the way you
determine iodine-131 in a sample, I think.
DR. POWERS: That's my impression.
DR. KRESS: Plus, you might add a little decay if
it's seven hours.
MR. HAYES: This is a curve of the Adams and
Atwood data, which is without the factor of three. You can
see that the highest spike that they had was approximately a
value of 4,000 and it occurred at around
ten-to-the-minus-three microcuries per gram.
You can look and you can see this is a spiking
factor one and you see that there are data points which show
a spiking factor of less than one.
If you look at the data from about
ten-to-the-minus-two, you can see that most of the points
are a value below, roughly, I'd say, 700.
DR. KRESS: The points below the line imply that
when you undergo this event, that the rate of release of
iodine from the pins decrease over what it was normally.
MR. HAYES: And you're going to ask me why that's
the case, and I don't know have an explanation, whether
that's just the data.
DR. KRESS: It could be data uncertainty, you're
right.
DR. POWERS: When I look at this plot, it seems to
me my first reaction is, gee, I believe I could run a
straight line through these data on a log-log plot, and when
I think about doing that, coming up and then using that
straight line to make estimates on the spiking factor as a
function of the RCS initial activity, I put error bounds
around it for what the error in the estimate would be.
Because I run out of data at the one microcurie
per gram, I would think that those error bounds would turn
up pretty dramatically.
Have you made such a plot?
MR. HAYES: We did. What we did is when we got
done with the industry data, Adams and Atwood, we thought,
well, we think that probably the spiking factor is a
function of reactor coolant activity level and we can have
-- we'll have our contractor analyze this data and hopefully
we'll come up with a plot of spiking factor as a function of
primary coolant activity level.
And we couldn't get a good correlation and we
really thought that we were going to be able to do that.
When we went into it, we thought we would be able to do it.
DR. CATTON: But you could easily get a bounding
curve for this.
MR. HAYES: And we did and the bounding value that
we got, the 95th percentile value we got was 335.
DR. CATTON: That's a single number.
MR. HAYES: That's a single number.
DR. CATTON: 335 times --
MR. HAYES: Times whatever the release rate would
be at equilibrium.
DR. CATTON: How is that bounding? I can see one
here that's more than 1,000.
MR. HAYES: We didn't take the highest value. We
took a 95th percentile value.
DR. CATTON: So that means you weighted a lot of
these negative ones down here.
MR. HAYES: That's correct. That's correct.
DR. POWERS: When I look at these data and I stand
far enough back from the plot, I say, gee, it looks to me
like there are two populations of data here. There is a
population that comes along and includes these low ones and
there's a population above.
Did you look at the data set to see if there was any
indication there were two populations?
MR. HAYES: We asked Adams about that to see if he
had an explanation and he didn't have an explanation.
MR. HIGGINS: That's a different question. What
Dr. Powers asked is if you did it.
MR. HAYES: No, we did not.
MR. HIGGINS: Is there at least some consistency
at a given plant? This is across many plants. Do you have
multiple data points at a given plant?
MR. HAYES: We have multiple data points, but I
don't think we would have a sufficient amount to really say
that this would be reflective of a plant.
See, I think one of the problems you would have
with such an evaluation is because of fuel changes. A lot
of the spiking data that we have is pre-1980 data, in the
'70s, because that's when you had a lot of these events.
And it's really not reflective of the conditions that you
have now.
It will give you data, but it's not really
reflective of what we have today.
MR. BALLINGER: Has it been plotted as a function
of the delta P?
MR. HAYES: No, it has not.
DR. CATTON: The points that are below that line
are really kind of perplexing. I don't know how you can use
them if you can't explain them. Maybe it's an error in your
data processing. Maybe it's an error in your readings.
So how can you use it when you attempt to
calculate some kind of a factor?
MR. HOLAHAN: I would suggest it's not really
reflective of what we have today. If you think about what
it means to go to the left on that chart, you're dealing
with very small numbers and you're dealing with the ratio of
small numbers, and I would expect the uncertainties to get
larger.
There are two things going on. I think Dr. Powers
suggested that there is less data as you go to the higher
numbers, clearly, but the numbers are harder to calculate as
you go to the left, so the uncertainties get larger.
DR. CATTON: I understand that, but normally when
you look at a plot like this and you're trying to figure out
what's going on and you want to come down on the safe side,
if the bottom data points which pull the number down can't
be explained, it seems to me they should be eliminated until
they can and not be a part of the process that leads to the
number.
MR. HIGGINS: Or call them, call them spiking
factor of one.
DR. CATTON: You can weight them. I mean, there's
all kinds of techniques for dealing with these things.
MR. HOLAHAN: But recognize that this is a log
plot and they don't change the answers by all that much.
DR. CATTON: I see that, I see that.
MR. HOLAHAN: I think you should have the same --
DR. CATTON: I have a straight line that
beautifully --
MR. HOLAHAN: You should have the same question
about the value of 4,000, which has a large uncertainty on
it.
DR. CATTON: It depends what you're trying to do.
You would ask the same question about the points that sit
way up by themselves, except that they're more important.
MR. BALLINGER: But there's very few instances
where you would just arbitrarily eliminate data points, but
one of them would be if it's simply non-physical, it can't
happen. Now, is that the case here with the ones that are
less than one? Is that a non-physical situation? That is
to say, it just can't happen.
MR. HAYES: I think Gary Holahan's point is a good
point. Remember that the spiking factor has a numerator and
a denominator and the values associated here are very low.
MR. HOLAHAN: It may be non-physical, but it's
giving you some insights as to the measurement of
uncertainties.
DR. POWERS: If we were to put error bars on these
data points, about how big would they be?
MR. HAYES: Since I didn't take the measurements,
I really couldn't answer that. I'm sorry.
DR. BONACA: It could be, in effect, related to
the type of transient that is taking place. For example, in
a steam line break, you have a depressurization in a system,
where probably, in the short term, you have almost
equalization, you could have, between system pressure and
internal pressure of the fuel rod.
MR. HAYES: It's important to understand that each
of the -- none of these events involve a main steam line
break.
DR. BONACA: But you have a scram. All right.
DR. KRESS: These are all team generator tube
ruptures?
MR. HAYES: Well, there are a couple of steam
generator tube ruptures, but most of them are transients.
They do not involve a steam generator tube rupture.
Fortunately, we have not had that many steam generators, or
we might be dealing with a different issue.
MR. BALLINGER: Which points on here deal with the
steam generator tube ruptures?
MR. HAYES: I would have to go back to the
original Adams and Atwood. I couldn't tell you at this
point.
DR. POWERS: It would be a substantial chore, even
if you went back to it, sorting through his table to figure
out which points are which.
MR. BALLINGER: But valuable.
MR. HAYES: I think you have the Adams and Atwood
and Adams and Sattison articles from Nucleonics Week, I
think, or some --
DR. POWERS: Nuclear Technology.
MR. HAYES: Nuclear Technology. And I think he
does list in those documents which ones are steam generator
tube rupture events.
DR. POWERS: Then when you find them and you go in
and try to find the point on the plot, it's a chore.
MR. HAYES: This is the same data multiplied by a
factor of three and that value, you can see, takes you
around 10,000.
Okay. The conclusions by Adams and Atwood, you
know, in many respects, similar to the conclusions of his
previous article. In other words, a spiking factor could be
reduced substantially, he believes, by a factor of 15 and
still be conservative.
Again, it's important to point out that we're
talking at the one microcurie per gram area. That's what he
was addressing. I don't want to give you a misleading
impression. That's what he was focusing on.
And the large spikes occur not because the
post-trip release is large -- in other words, the numerator
is large -- but, rather, because the steady-state release is
low -- in other words, the denominator is small.
And that the spiking data that he had was really
representative of a steam generator tube rupture rather than
a main steam line break.
When we did our reassessment of the spiking
factor, we reviewed the data without the factor of three
and, as I mentioned previously, we thought we could assess
it and come up with a relation as a function of reactor
activity level.
DR. POWERS: Let me understand a little better.
Adams had a reason for the reactor of three. He did it not
out of any capriciousness. He did it because he --
MR. HAYES: He wanted the bounds.
DR. POWERS: But he had a rationale for choosing
three. I mean, it wasn't a number that he plucked out of
the air, I don't think. I think he made arguments about
sampling and stuff like that, as Dr. Kress pointed out.
Have you rejected all those arguments?
MR. HAYES: No, we didn't reject -- in reality, it
probably doesn't make a difference, because we -- his
argument was at the value of one, the value of 500 was
conservative.
DR. CATTON: Well, it is, but from that chart you
put up there, it looked to me like it's only maybe a factor
of five. Put that chart back up, the first one you had up
there.
I think that's one that's on the right-hand side,
isn't it?
MR. HAYES: Right.
DR. CATTON: How can you get a factor of 15? Do
you take the mean through all those points? Is that what
you're doing? The 95 percent level, just looking at the
graph, looks to me like it's about 90. Maybe I'm not
eyeballing that quite right.
MR. HAYES: You're asking me to justify his
conclusion and I'm not going to --
DR. CATTON: That's fine. But you're asking me to
accept 335 based on him saying that it was 15. I'm just
turning that around. I think it's fair. Don't you?
MR. HAYES: That's fair.
MR. HOLAHAN: Well, wait a minute. I don't want
to be too fair. I think the 335 is an analysis of the data,
independent of whether someone else concluded that 15 could
be drawn out of that data, and we've shown you the data.
DR. CATTON: Well, I'm still bothered by the fact
that you can't explain the data points, yet you want to
treat them all as equal, and it seems to me that if you
can't explain the low ones, which weight the number, then
you ought not weight the final result by those points.
DR. POWERS: On the other hand, if I look at the
data and I said, gee, I was going to take a 95 percentile
and I threw away everything that I call on my second
category, 335 isn't going to be far away from what the one
microcurie per gram.
I mean, it's not going to be far away and I'm not
going to be -- if I'm in a bounding sense, for that one
point, I'm not going to feel bad. In fact, I'd probably
feel guilty about picking a number that high.
DR. KRESS: If you did multiply the number by
three, however, you would multiply that 300-and-something by
three, basically.
DR. CATTON: But, Dana, they're going to run out
at .001. Now, we haven't gotten to a number for .001 yet.
Well, it's 335 times .001 as the spiking factor. Where are
you going to fall on this graph when you do that?
MR. HOLAHAN: The .001 never shows up in a
regulatory analysis, because there are not and there will
never be any tech specs as low as .001.
We're only talking about the portion, that curve
between one and a few tenths.
DR. CATTON: I take .1 and multiply it by 335,
what do I get? 33.5. I'm down to the middle of all those
data points. That doesn't look very conservative to me.
DR. BONACA: To the right.
MR. HAYES: At .1, you're right here.
DR. CATTON: But if I multiply .1 by 335, I'm
going to be just a little bit above that ten over there.
That's a log scale.
DR. POWERS: Why would you want to do that
multiplication? Well, they're going to use the tech spec of
.1. They still won't be multiplying the spiking factor.
DR. KRESS: They won't be multiplying this number.
They're multiplying some other number.
DR. CATTON: What number do you multiply 335
times?
MR. HAYES: You multiply by the release rate
associated with this particular activity level in the
coolant. In other words, for example, we don't presume that
the activity level is now 33.5 because we multiply 335 by
this value. No.
What it is is release rate and you build up in the
activity in the primary coolant.
DR. POWERS: He's going to take a removal rate and
a release rate and find a steady -- choose that release rate
such that it matches the steady-state value of .1.
DR. KRESS: That's what they're going to do.
DR. POWERS: And he's going to multiply that
release rate by a number.
Now, if the steady-state value were one microcurie
per gram, we know it's 335 he's going to multiply it by.
MR. HAYES: For a steam generator tube rupture.
DR. POWERS: For a steam generator tube rupture.
MR. HAYES: Not for a main steam line break.
DR. POWERS: Exactly.
MR. HAYES: I think it's important, with respect
to the point that you raise, is that since we did not take
the data, it's inappropriate, I think, for us to throw any
of these points out, because we have -- what is the basis?
I mean, you can go argue from either stage.
DR. CATTON: It depends what you're doing. If
what I'm looking for is just a mean curve through the data
of some kind, fine, but this is the safety business. If you
don't understand it, how can you use it to bring the number
down.
MR. HAYES: But it's a safety business, but at the
same token, we have to be somewhat realistic.
DR. CATTON: There's nothing wrong with being
realistic if you understand it. If you don't understand it,
it seems to me it's inappropriate.
MR. HAYES: But it is also inappropriate to throw
it out because you don't understand these points, as you
don't understand these points.
DR. CATTON: Depends whose side you're on. If
you're putting a barrier between me and the plant, I don't
want you reducing anything that you don't understand.
DR. POWERS: That's one of the reasons we wanted
to explore the phenomenology a little bit is that the lower
points do seem to be physical, whereas the upper points,
high though they may be, at least are not inconsistent with
the argument on why there is a spiking factor all together.
MR. BALLINGER: But the physics, the description
of the phenomena that you gave right off the bat matches
what you would see, but the low points, the ones that are
less than one, don't match it at all.
MR. HAYES: I don't disagree.
MR. HOLAHAN: I think if you were to put error
bounds on the measurements, real error bounds on the
measurements, they would, by definition, on the lower end,
be large enough to be above one, and then you ought to put
those error bounds on all the data and you'll see, as you
move to the left, the error bounds get very, very large.
DR. POWERS: If that's true, the error bounds
would cross multiple decades. I think if you were doing a
regression of the data, that you would take that into
account.
Those data points, they are high and low, would
count very little, which is fine because as you said, the
only part of the curve that you're really interested in is
one, maybe .05, at most.
MR. HOLAHAN: Right. Personally, I see the data
in different ways. In my mind, the data between .1 and one
is reflective of what we might possibly use in the
regulatory process.
By the way, all of those data points are above
one. The rest of the data, in my mind, is sort of a
demonstration that, in fact, iodine spiking is a real
phenomenon, that it is seen, and you have a lot more data
points.
But I don't think it tells you a lot about the
range and the statistical analysis on the right-hand curve.
DR. POWERS: I think that's the point that's going
to be -- we're going to discuss a lot of things here, but I
think that's the most distressing thing; that let's confine
ourselves one to .1, because that's all we're going to use,
and say -- I'm looking at a log-log plot here and there is
scatter on a log-log plot.
That suggests to me that we do not have the really
operative physical variable being plotted. Much of these
data, similar to a steam generator tube rupture, are not
steam generator tube ruptures. There's some other accident.
There is something affecting those data over a
decade in value other than just the activity in the coolant.
In fact, we think the activity in the coolant can't possibly
have any relationship to the mechanism giving us a spiking
factor, second order, maybe.
How then can we be confident that a bounding that
he does even at 335 really represents a bound on what
happens in the steam generator tube rupture, if there is
some other variable that is really controlling that spiking
effect.
Otherwise, you've got a Poisson problem here.
These happen to be the 25 data points that you've actually
measured, which may accidentally not hit the particular
value of the controlling variable that gives you high value.
DR. CATTON: I think if you cut the data off where
you suggest and you were to redo it, you'd get a bigger
number than 335.
DR. POWERS: You may be right.
MR. HAYES: Everything high seems to be at about
.05 on down.
DR. CATTON: Any time you have two decades of
spread, you really ought to go back and replot it, and if
you can't go back and replot it, you put a curve over the
top of it. You can't explain it and to do anything else, I
think, is an in-road on the margin of the plant, because you
don't know what point is there for what reason.
MR. BALLINGER: I think the idea -- the contention
that you may not have the right variable is probably the one
that's closer to the mark, that you just don't have a handle
on the phenomena.
MR. HAYES: There is no question that the iodine
spiking has not been pinned down specifically.
MR. HIGGINS: Isn't it really a plant-specific
thing? A fuel, actually a fuel and a core load specific
thing. It seemed like the amount of iodine that leaks out
is very much related to the type of defect you have in the
fuel and the specific type of fuel.
So what you're seeing here is a demonstration over
many cores, over many years at many different plants.
MR. HAYES: Exactly. And I think I made reference
to the fact that most of this data comes in pre-1980.
DR. CATTON: It sort of presents an option,
doesn't it? You either accept it or you go back and do it
again and do it right.
DR. POWERS: I think there's another alternative
here that I'd like to understand a little better. You
mentioned the Iglesias-Lewis report and I think there's also
an EPRI empirical report and in both of those reports, they
look at some notable events and they plot predicted curves
versus the iodine concentration, I believe, and they're
remarkable, actually.
I mean, the closeness with which they get just
really amazes me. And there's a substantial time variation
in what they're comparing against. Can you tell me why
those empirical and mechanistic models were just cast out in
favor of this data?
MR. HAYES: They weren't cast out. What we did is
we had a contractor, INEL, evaluate those models and they
thought that the -- after looking at the empirical model,
they thought that probably the model to really look into and
expend their resources on was the Lewis and Iglesias.
And what happened was they evaluated the Lewis and
Iglesias model and what they did is, for example, if the
event was at Prairie Island, they tried to take that model
and utilize the predictiveness to determine what the spike
would be for Prairie Island.
And when they did that work, you couldn't take it and
predict it for Prairie Island, San Onofre, Watts-Barr or
whatever. They couldn't come up with it.
So I think the way I related to it is similar to
when you're in a lab and you're trying to come up with a
relationship. You do it after you have the data versus
before.
So a priori, they couldn't come up with a
predictive tool using the Lewis-Iglesias. Adams, when we
discussed this with him, he says he recommended, if we could
modify the Lewis and Iglesias model, that that would be an
approach to go and, therefore, maybe we could come up with a
predictive tool, and that was one of the things we raised
with industry, but industry did not choose to pursue it.
DR. CATTON: If you did fit -- if you fit a curve
over just the two top data points, I bet you'll get a factor
that is much less than the 300 when you're operating greater
than .1. You're going to get about 120.
MR. HOLAHAN: Log scale, I think you get about
200.
DR. CATTON: Just by bounding the data. It's a
nice logarithmic curve.
MR. HOLAHAN: But let me suggest that this is one
piece of a design basis dose calculation which has lots of
other conservatisms in it, like 95th percentile meteorology
and lots of stuff.
And if we're not careful, we'll end up making
every step in every calculation so conservative that the
answers are meaningless and we're in danger of convincing
ourselves to do things which risk-informed regulation, in my
mind, is trying to get us away from, which is to put lots of
attention on things that aren't necessarily important.
So when you say we could be more conservative by
drawing the lines differently, that can be true, but that's
not always the best regulatory approach.
DR. CATTON: You could include uncertainty in
whatever fit you put on this, too.
DR. POWERS: Fortunately, our concern right now is
really going contention by contention and not designing a
regulatory process.
MR. HOLAHAN: I've tried to preserve a concern
that we ought to have an overall safety perspective and the
regulatory approach in mind when we deal with each of these
issues.
MR. HAYES: Okay. What did we conclude with
respect to iodine spiking? Well, we concluded that it was
not a function of the reactor coolant activity level. We
did envelope the value, as we mentioned, and we came up with
a value of 335 and, again, that is what we considered to be
representative of a steam generator tube rupture and not
representative of a main steam line break. I think that's
important to understand.
What are our conclusions with respect to the
spiking associated with the main steam line break? Well, we
concluded that you cannot extrapolate the spiking data from
the data we have to a main steam line break, but we can do
some sort of an assessment. We believe that there is a
linear relationship between the differential pressure rate
and the resultant iodine activity or release rate.
Now, because of the pressure, differential
pressure associated with the main steam line break being
maybe a factor of two or three greater than for a steam
generator tube rupture, if you consider that to be two or
three times the factor for a steam generator tube rupture,
you also can say, well, maybe the expression is a
quadriture. So instead of a factor of two or three, the
value is four to nine.
I think we feel confident and we've had some
discussions with Adams that it's reasonable to assume that
the expression is no higher than a quadriture.
DR. POWERS: I guess I would understand that
better if I understood the first line. You said there was a
linear relationship between the delta P and the resultant
iodine activity level or release rate.
Why do you have that confidence?
MR. HAYES: I think because, you know, it is the
delta P that causes the spike, it forces the fuel out of the
gap into the reactor coolant.
DR. POWERS: That delta P arises because of the
vaporization of water on the hot fuel.
MR. HAYES: The delta P originates from the change
in the fuel temperature and pressure and the primary coolant
temperature and pressure. Then when you get to the main
steam line break, the presumption is that you have it with a
loss of off-site power. So you're going from the primary
side to the secondary side and you have a direct release
associated with the main steam line break.
You don't have the cover associated with it like
you do with a steam generator tube rupture.
DR. KRESS: The delta P rate you're talking about
here is the rate of change of pressure on the primary system
in the vicinity of the core. I don't know what delta P
you're talking about here.
DR. BONACA: I had the same question.
DR. KRESS: It's a del P, so it's a rate of change
with time.
MR. HAYES: It's a change with -- it's not a rate.
It's a delta P, period.
DR. KRESS: But what delta P is it?
DR. BONACA: But is it a delta P between the
primary system pressure and the pressure inside the fuel
rod, or is it the pressure --
MR. HAYES: It's the pressure between the primary
side to the secondary side.
DR. BONACA: Explain to me what effects that has
on the fuel cracks opening and accepting more water inside
and vaporizing and then exiting from the fuel rod. I don't
understand how the delta P between primary and secondary
side is affecting that. I'm trying to understand it.
MR. HAYES: For the main steam line break, the
faulted steam generator, that which experiences the break,
is considered to be at atmosphere.
The leak which we have is then from the primary
side to the secondary side. We get a rather rapid
depressurization on the secondary side. So you set up
between the primary and secondary a higher differential
pressure.
Then the primary pressure goes down faster than it
would for a steam generator tube rupture. So the pressure
between the fuel and the primary side is a larger -- is a
quicker and a larger delta P.
DR. BONACA: Between the fuel and the primary
side.
MR. HAYES: Between the fuel and the primary side.
DR. BONACA: Okay. I understand.
DR. HOPENFELD: Part of the study came out of the
Adams study that we were going to look -- take one step
further, and that is, if you have to look at that data in
relation to the depressurization, what happened during this
transient and if you look at some of the data or some of the
reports on that subject, it shows that you can very, very
low release rate; in other words, you have very, very few
defects.
Then you find in those cases, all the activity
comes out at the end of the transient, where you really have
to depressurize is very, very high, and that suggests that
when you have a steam line break, you have much more release
because of the high depressurization.
In other words, there is a relation between the
number of defects you have and how fast you depressurize the
system. But nobody really has taken the time to look into
that beyond that point.
I can provide you the conclusion of a very lengthy
report that was generated at Westinghouse.
DR. KRESS: I'm still confused by that first line.
Tell me again what the delta P is. Is it between the
primary and the secondary?
MR. HAYES: Okay. It starts out between the
primary and the secondary and the secondary -- okay. With
the main steam line break.
DR. KRESS: Yes, and it has a fixed value at
normal operation.
MR. HAYES: Normal operating. Then with the steam
line break, you go to atmosphere.
DR. KRESS: That's right. So that --
MR. HAYES: That reduces it down --
DR. KRESS: That reduces it. It increases the
delta P.
MR. HAYES: It increases the delta P, and then you
have --
DR. KRESS: Are you talking about the rate of that
increase?
MR. HAYES: It increases both the rate and the
net.
DR. KRESS: The net and the rate.
MR. HAYES: Right. Because you're going -- like,
you're comparing it to a steam generator tube rupture.
DR. KRESS: What in the world should that have to
do with the rate at which you extract iodine from the fuel?
DR. BONACA: The only way I would see it would be
that the primary system pressure is also dropping fast.
DR. KRESS: That doesn't say that, though.
DR. BONACA: That's right. So that's why I was
confused. I mean, I believe that -- I believe that the
difference in pressure between internal pressure of the rod
and the primary system, it's varying rapidly.
DR. KRESS: Then I could see it would have a
marked effect on it, but this difference between primary and
secondary.
MR. HAYES: But why is the primary varying? The
reason the primary is varying is because you have a change
on the secondary side, also.
DR. BONACA: Sure, but --
MR. HAYES: So you can't separate one from the
other.
DR. BONACA: I understand that, but that's a
driving force by which you are having a changing
relationship between primary system pressure and internal
pressure in the fuel rod, and that's the mechanism by which
you would expect to have release of iodine activity to the
higher level.
DR. KRESS: Let's say, for example, that you had
no leaking steam generator tubes at all and you had this
break in the secondary side, you get a marked change in this
delta P and the rate may even vary with time, does nothing
at all to the primary system because it's just sitting
there.
There's no leak coming out. It's all driven by
the secondary change and should have no effect on the iodine
spiking at all or the iodine levels. That's why I don't
understand the statement.
DR. BONACA: The only thing you have is scram and
that will have some changes in system pressure.
DR. KRESS: If you scram, yes, but that's --
DR. BONACA: That doesn't have the same effect.
So the driving force, seems to me, it would be on the effect
that we're looking for would be the difference between
primary system pressure, driven by --
DR. KRESS: Driven by that, yes.
DR. BONACA: -- the steam generator break. I
understand that. And the internal pressure of the fuel rod.
MR. HAYES: But it will change your cool-down
rate, though. You say it will not an effect, but it will
have, because it will change your cool-down rate, because
you're using the intact steam generators to cool down the
primary side.
DR. KRESS: It might have some effect, you're
right. That's pretty --
DR. BONACA: What is the average internal pressure
of the fuel rod?
DR. POWERS: During operation?
MR. BALLINGER: What's the internal pressure?
DR. BONACA: Yes.
DR. POWERS: An intact one?
DR. BONACA: The fuel rod in a -- well, I'm
talking about an intact one.
DR. POWERS: It would run about 100 atmosphere.
DR. BONACA: That's right.
DR. KRESS: So you look at that delta P that you
normally get with the steam generator tube rupture and then
you look at it, once you get to the main steam line break,
and that's where you get this factor of two to three.
MR. HAYES: Right.
DR. KRESS: If you square that, you get four to
nine.
MR. HAYES: That's correct. So what we concluded
with respect to the main steam line break, and, again, this
was in conjunction with the steam generator rule and
addressing the issue of conservatisms in our evaluation, was
that for a main steam line break, we thought that there was
probably an uncertainty factor of ten.
Adams indicated, for a steam generator tube
rupture, there is a conservatism, he believed, was a factor
of 15.
DR. CATTON: This is the 15 I don't like to use.
MR. HAYES: This is the 15 you don't like to use,
right.
DR. POWERS: But what I don't understand is that
you went through and you looked at the data and you said,
okay, I'm not buying this factor of three bounding and you
looked at it and you came up and you said 335 versus 500,
which is not a factor of 15.
How come now, all of a sudden, you grab a hold of
Adams' factor, which nobody understands?
MR. HAYES: Because we incorporated more than just
that particular factor.
DR. POWERS: I kind of wish you would tell us what
those things were, because otherwise, this third statement
down here is just going to cause me to --
DR. KRESS: To go ballistic.
DR. POWERS: Yes, ballistic, because that is
definitely not the way you handle combined uncertainty.
MR. HAYES: The imaginary axis.
DR. POWERS: I mean, this is not even -- would not
be acceptable in anything that I can think of, where you
take two things that you don't understand at all and say
they offset each other.
MR. HAYES: We concluded that the doses associated
with a main steam line break are typically on the order of
one to three rem, the ABLPZ, control room.
There is also a parametric analysis which we have
done to demonstrate what the spiking factors would have to
be in order to get a dose which would exceed Part 100. The
criteria -- remember, the criteria associated with this is
for a dose of 30 rem, not 300, but for 30.
Even if you took the factor of 335, which is a
value, what, of 1.5, roughly, 1.6, 7, and take this factor
of ten, in our opinion, you are still within the uncertainty
that we had and still within the margin.
DR. POWERS: I can buy arguments that go that
direction. I just can't buy this slide. I find no
technical justification for the third sentence. That's the
problem. A bounding operation that came in and said, look,
the sensitivity to my dose calculation or the spiking factor
is smallish. I have to get a very big value that seems to
strain quadrulity to approach the limit, so I'm going to
leave it at 500. I'd say, okay, fair enough, that's a
decent argument, and go on.
Something that's just orthogonal to the treatment
of uncertainties, this gets me excited. You're more relaxed
than I am.
DR. KRESS: I just don't show it.
DR. CATTON: This last statement about should
remain 500, that sounds like you completely -- you have
either written off or ignore or don't believe that the main
steam line break can have some impact on the internals of
the steam generator, because it's certainly -- at least from
what we heard yesterday, it's most likely going to cause
more leakers.
Shouldn't that be factored in somehow?
MR. HAYES: It's already included when we
postulate that those tubes which have voltage-based
criteria.
DR. CATTON: But that's before. If that's what
you're doing, what you're saying is that the main steam line
break does not cause any disturbance inside the steam
generator. The tubes are going to stay exactly the same.
MR. HAYES: No. What we have presumed is that we
have presumed that those tubes which have voltage-based
criteria, those cracks are going to open up and they're
going to leak at a defined rate.
In other words, every tube to which that criteria
has been applied is assumed to leak.
DR. CATTON: But that leakage correlation was
derived from tested tubes that have not been subjected to
the main steam line break. Am I missing something?
MR. STROSNIDER: This is Jack Strosnider. I'm not
sure that Jack Hayes heard the discussion yesterday on
dynamic effects and some of the Surry and other experience.
We're going to talk about that later today.
DR. CATTON: But it impacts this.
MR. STROSNIDER: Well, it may or may not. I think
apparently you've reached a conclusion that those events
are, in fact, going to cause additional damage to the tubes
and we heard the discussion yesterday.
We've put this into the GSI process and we're
going to hear about how that's being looked at. If, in
fact, we conclude that those dynamic effects are going to
have that sort of effect on the tubes, then we'll have to
deal with it, but I don't think we've seen that concrete
evidence to this point.
There's certainly, in my mind, some things that
ought to be followed up on with regard to was there a
post-event inspection of the steam generators, what did they
actually find, the tubes that were indicated as having
degradation in yesterday's presentation, it wasn't totally
clear that that was a result of the event.
So I think it's an issue that needs to be looked
at and when we determine whether it really has an impact on
the tubes or not, then we need to come back and deal with
it. But at this point, you're right, the model does deal
with pressure-induced leakage, and we'll talk about that
later today, too.
DR. CATTON: Just if you had another line down
there that said that this is an assumption. See, you're
basing this on the assumption that the main steam line break
does nothing to the internals of the steam generator and at
this point, that's your assumption. I can read it, I
understand it, we can go on.
MR. STROSNIDER: Right. And we've got some new
information here that needs to be assessed and we're going
to do that.
DR. CATTON: That's right. But that isn't what he
was telling me.
MR. STROSNIDER: And like I said, coming back to
my some of my introduction this morning, we need to go
through and hear all the issues and look at how they all fit
together, and that's a fair comment.
DR. CATTON: That's fair enough. That's fair
enough, but I would liked to have seen another line on that
viewgraph.
MR. HOLAHAN: Remember, it's more than assumption.
It's a requirement. Plants are licensed and that licensing
basis includes looking at main steam line breaks and tube
ruptures, but it doesn't include main steam line breaks
damaging tubes.
And if we find some basis for thinking that that's
true, we'll have to deal with it, but the licensees will
have to deal with it because that will be inconsistent with
the current requirements.
MR. HAYES: Now, getting to the assessment of the
DPO concern, we did a parametric analysis and we presumed --
we used, as the base case, a three-loop Westinghouse plant
and we did the analysis consistent with a standard review
plan 15.1.5, which is the main steam line break.
And we presumed tech spec value one microcurie per
gram. The primary to secondary leak rate was 150 gallons.
This should be corrected on your slide to per steam
generator. And this value is, instead of 1290, should be 11
-- I think it's 1140.
I didn't conclude the two steam generators, just
the one. It really doesn't have an effect -- it was
included in the analysis, but not in the slide.
Spiking factor was 500 and what we did is we
calculated the releases for zero to two hour and zero to
eight hour time period.
Some other critical assumptions that we had, we
assumed that it was eight hours for the faulted steam
generator to be isolated. All primary and secondary leakage
was assumed to be released directly to the environment. We
did not presume, for example, for the intact steam
generators, any partition factor.
It really doesn't have a whole big effect on it,
an assumption, it's probably in the third decimal place or
third significant number.
Spiking was assumed to occur for the duration of
the accident and --
DR. KRESS: What does that mean? Does that mean
that you kept the rate that you calculated constant over the
whole eight hours?
MR. HAYES: Yes, for the whole eight hours. You
think that's conservative?
DR. KRESS: I think that is.
MR. HAYES: And here is another big assumption.
We assume that the releases associated with zero to two
hours and zero to eight hours equated to a 30 rem thyroid
dose. Now, as I've mentioned before, the typical values we
see are somewhere between probably one to five or six rem.
Then we did a parametric analysis. We assumed
three different primary to secondary leak rates, ten, 35 and
100 GPM. We assumed five different primary coolant activity
levels. I think your slide has an error, a typo. This
should be ten-to-the-minus-two. I think in your slide it
has ten-to-the-minus-three. But these are the numbers we've
presumed.
DR. POWERS: Gary tells us that .005 is not and
will never -- ever going to occur.
MR. HOLAHAN: That's right, and the reason is not
because I'm against low primary coolant activity. It's
because I'm against allowing leakage rates higher than 100-
GPM, which is what the implication is. And the point is the
design basis calculations may come out with a reasonable
dose, but the severe accident implications I think we would
find unacceptable.
I think we will cover those sort of issues later.
MR. HAYES: I may need to go put in some
clarifying information with respect to that. I think we
need to probably check what the value was associated with
Byron and Braidwood, because I think Byron and Braidwood had
a number which was either at 100 or slightly greater and the
value may have been down below .05.
But that's something we need to check on, because
they have subsequently -- I think that was for one cycle and
they subsequently replaced their steam generators. That
would have been Byron Units 1 and 2 -- excuse me -- Byron
and Braidwood Units 1. Unit 2 did not change steam
generators.
MR. BALLINGER: These leak rates are prior to the
event or during the event?
MR. HAYES: This is during the event.
MR. BALLINGER: During the event.
MR. HAYES: During the event. What you can
consider this to be is your accident-induced leakage.
DR. KRESS: You're calculating the spiking factor
you would have to get to get a 30 rem dose here.
MR. HAYES: Well, it was presumed that the release
that we calculated for zero and two hours gave you a 30 rem
dose. It did not. It did not, but we presumed that.
DR. KRESS: Okay. And you back-calculate those to
find out what the spiking factor would --
MR. HAYES: Right, exactly.
DR. KRESS: -- would give you that.
MR. HAYES: Right. Here is what we did, for
example. Let me walk you through a couple cases. We
presumed RCS activity was .5. We took a ten GPM leak rate.
In order to come up with the same release rate, we would
have to have gotten a spiking factor of 86.3.
So for example, if you wanted to operate with the
ten GPM leak rate, based upon our criteria, you would have
to go down to a spiking factor of 500 before it would be
acceptable to the staff.
So you would go down to the -- the reactor coolant
activity level would have to be reduced from .5 to .1.
DR. KRESS: I understand what you're doing now.
MR. HAYES: That's what we were doing. Okay.
Now, the premise is at low rates of reactor coolant activity
level, that the values would exceed Part 100 doses, that the
spiking factor would be greater than at 5,000, and you would
have to exceed Part 100 doses at those levels.
If you go down to ten-to-the-minus-two, you can
see that even at a release rate of 100 gallons per minute,
the spiking factor is at least 500 or greater. If you went
down to ten GPM, you're at 5,000. That's for a dose of --
that's presuming that your dose was 30 rem.
So to exceed Part 100, it would be ten times this
value. You'd have to have a spiking factor ten times this
value. It would be 5,000, 6,000, 51,000, that would have to
be the number.
DR. POWERS: In order to do these calculations,
you had to calculate what the steady-state release rates
were.
MR. HAYES: That's correct.
DR. POWERS: Do you know what those numbers were?
MR. HAYES: I would have -- I could provide them
to you. I'd have to look at what they were.
DR. POWERS: I think I'd be interested in seeing
an example calculation.
DR. KRESS: In order to do that, you would have
had to assume something about that the size of the primary
system and the capacity of the cleanup system, did you use
some sort of generic numbers for those?
MR. HAYES: I did a specific plant example. I
took --
DR. KRESS: This was a specific plant.
MR. HAYES: Yes. What I did, if you go back to
the base case, I took a particular plant, a three-loop
Westinghouse, took the numbers I had for that and then I
adjusted it to these situations. So it was an actual
example.
MR. BALLINGER: What happens if you try 1,000
gallons a minute?
MR. HAYES: I think the point -- first of all --
MR. HOLAHAN: Your division director goes
off-scale.
MR. HAYES: You have a LOCA, you don't have
sufficient makeup. You start to get above now. I think
some of the people from the staff can answer that, but I
think if you get much above 100 GPM, you have a problem with
makeup.
MR. BALLINGER: Well, previous tube rupture events
have resulted in three, 400, 500 gallons a minute.
MR. HOLAHAN: Yes, but that's not what we're
talking about here. We're not talking about tube ruptures.
We're talking about acceptable post-accident leakage.
MR. HAYES: Because this is going directly to the
environment. You have no water level above your release
point.
I think this goes to your argument in considering
safety and Gary said that, hey, if we start to go to 100
GPM, he starts to get going off-scale.
I think realistically, what do we have to be
concerned about from an accident standpoint? Isn't our weak
link the steam generator? So that when you start getting
into these areas, you have a concern.
Most of the people we're dealing with are in the
ten to 35 GPM range. That's where the numbers are at. Now,
we did have one case, yes, where Byron and I think
Braidwood, or maybe just one of them, was in that ballpark
for either a portion of a cycle or one cycle before they
replaced their steam generators.
MR. HIGGINS: And those numbers in the second
column are the ones you're going to talk to us later, about
how those post-accident numbers are derived for the main
steam line break.
MR. STROSNIDER: You're referring to the leakage
values.
MR. HIGGINS: The second column there.
MR. STROSNIDER: Yes, and we'll explain how that's
derived as part of the generic letter process.
DR. BONACA: The DPO makes the contention that the
ultimate repair criteria results in a change to the design
basis event that will cause to have a steam line break with
leakage rate exceeding this number, right?
MR. BALLINGER: That was the point I was going to
get at.
DR. BONACA: Right.
DR. KRESS: You said later on you were going to
discuss that point.
MR. STROSNIDER: Right. As part of our
description of Generic Letter 95-05, we'll explain how those
leakage rates are calculated, if you will, design basis
leakage rate.
MR. HAYES: Again, reiterating, at .01, we're
talking about a spiking factor of 5,000 to 51,000 in order
to exceed Part 100 doses and for lower reactor coolant
activity levels, you can see the number gets even larger.
Again, this is the Adams data that we spent a lot
of time looking at and discussing and the maximum value is a
value of 10,0000.
And, again, we presume that the releases
associated with those two time periods were at the 30 rem
limit, which they're not. They're another factor at least
three to ten lower.
The conclusions from the table; obviously, for
some combinations of leak rate and primary coolant activity
levels, it would require a spiking factor of less than 500.
However, for ARC amendments, this would necessitate reducing
the primary coolant activity levels and that's what they do.
The spiking factor, because we use 500 in our
calculations, the actual spiking factor would have to be at
least 5,000 for Part 100 limits to be exceeded, and that's
based upon the 500 times the 300 rem Part 100 limit divided
by the 30, which is our use.
And then for primary coolant activity levels of
.01 or ten-to-the-minus-two microcurie per gram, the spiking
factor would be 5,000 to 51,000.
DR. POWERS: Let me ask a question on this.
Suppose in my criterion, I did not take the Part 100 limits,
but I took GDC-19.
MR. HAYES: Which is the third.
DR. POWERS: Control room habitability effect.
MR. HAYES: You want the -- the answer to your
question is this. It's what it already is right here,
because that's what it's been based upon, 30 rem thyroid,
same as GDC-19. So at the ten-to-the-minus-two, we'd be at
the 500.
DR. POWERS: But if I look at the Adams data or
ten-to-the-minus-two, I can find numbers that exceed 500.
MR. HAYES: Yes, you can. Yes, you can. But you
don't -- you see them exceeding -- this is at three times.
You don't see them -- let's see. Well, these are the ones
over 1,000 right here. That's with the multiplier of three.
There's not a whole lot of points.
And, again, look at where we're at for 1,000.
Even at .05, we're at ten GPM. We're at 35 point for
ten-to-the-minus-two.
MR. HIGGINS: So was the intent of this to justify
the 500 spiking factor for main steam line break? Is that
the intent of this example?
MR. HAYES: No. The intent of the example was to
address the DPO and to say that, hey, at these particular
low reactor coolant activity levels, the source of iodine
that you have in coolant, both in terms of the initial
coolant and then the release rate, is not significant enough
to put you over Part 100 limits, and that even if you did,
for example -- let's say our factor of 500 is wrong.
I think you can see from the numbers here, at
these various leak rates, you would have to have a
significant spiking factor in order to exceed just the 30
rem.
DR. POWERS: I think we have a significant spiking
factor. Suppose that I go to the Adams data. Suppose that
I, say, factor of three that he multiplies things at, but
suppose that I do buy the linear hypothesis presented
earlier on the delta P.
I don't even ask for the quadriture process. I
just use the linear. Then I'm back onto this slide here.
MR. HAYES: Yes.
DR. POWERS: But I'm interested in complying with
GDC-19, which has just as much effect on me as Part 100. I
have to take that into account just as much.
DR. KRESS: And I see no reason not to accept the
factor of three. So if you use that and the other factor of
three --
DR. POWERS: Then you're in serious trouble.
MR. HAYES: I don't think you are in serious
problem. Look at this again. This is at the GDC-19 value.
This is at 30. You have presumed that these releases are at
the 30 rem, and they're not. These releases are not at the
30 rem. If you take a base case and some of you that come
from plants, you take the base case for these plants for
main steam line break, the value is probably no more than
one to two, three rem.
For example, we just did an ARC amendment, we're doing an
ARC amendment for Watts-Barr and it's a ten GPM leak. This
is instead of one. And the doses, the maximum dose is two
and that's at the LPZ. Control room is one.
Now, for a regular accident, the main steam line
break isn't your limiting primary to secondary accident.
You're steam generator tube rupture is.
So this number probably, if you actually went to
one, it would probably be even lower. But we have presumed
that. If you want to take -- okay. We're going to say we
have an uncertainty of ten. So what, you say the spiking
factor has to be 5,000 then. Spiking factor of 5,000.
Okay.
At ten-to-the-minus-two, you're still at ten GPM,
and look what happens when you go here.
DR. POWERS: Look, I'm not interested in looking
at it, because Gary has assured me I'll never go there.
DR. KRESS: I'm only interested up here at the .1,
the point level. And even at ten GPM --
DR. POWERS: I think you're in a world of hurt at
ten GPM. Now, if we factor in this additional thing that,
in reality, the doses for this particular event at this
particular plant at the control room is, as you say, three
to five rem thyroid.
MR. HAYES: Probably the maximum, yes.
DR. POWERS: It's not obvious to me we're out of
the woods either. That doesn't get you quite there either.
DR. KRESS: That I presume is taking account for
transport.
DR. POWERS: I think the reality is that when they
do this exact calculation for a particular plant as part of
the FSAR --
DR. KRESS: The atmospheric transport.
DR. POWERS: Nobody comes back and says I'm at 30
rem, they always come back and say I'm five or six and less,
.5.
MR. HAYES: Sometimes, in reality, with respect to
the ARC amendments, the limiting is the control room and
sometimes those values are high.
One of the reasons why we stuck to releases versus
calculating doses is because we threw out the atmospheric
dispersion and threw out, if you will, the control room
removal mechanisms. We thought that was a less biased type
of approach.
DR. KRESS: That's a good thing to do if it gets
you out of the woods, because you're all right, but we're
not so sure that gets you out of the woods yet, because
we're not sure the spiking factor might not be 5,000, for
example.
MR. HAYES: At this particular point in time, the
staff has accepted the spiking factor of 500. So, for
example, if we were at this level, at
five-times-ten-to-the-minus-two, and found at 35 we were not
at the spiking factor, we would have to go somewhere between
.05 and .01 and at 35 GPM, you'd be going from 283 to 1490.
So you're probably talking about .4, in that vicinity. That
would have to be our acceptance criteria and that would be
for 30 rem.
DR. POWERS: I think it may be safe to say we
understand what was done.
MR. HAYES: Okay. Our conclusions with respect to
the DPO concern, yes, we agree that spiking factors greater
than 500 can occur, but they are low dose equivalent
iodine-131 activity levels.
We don't believe in any case that the spiking
factor would be less -- would be greater than 5,000 and
based upon the parametric analysis we did, we believe that
if you were at an activity level of less than
ten-to-the-minus-two microcurie per gram, that a spiking
factor would have to be at least between 500 to 5,000 for
the base case releases.
If you look at the amount to exceed Part 100, it
would have to be ten times that or 5,000 to 50,000. At
primary coolant activity rates of ten-to-the-minus-two, the
primary coolant content is small and the equilibrium release
rate is small.
And as we mentioned, we don't believe the spiking
factors are greater than 5,000.
Are there any other questions? That concludes our
presentation.
DR. POWERS: Seeing no questions, I think I will
declare a recess until five after the hour.
DR. KRESS: I have just one question.
DR. POWERS: I think I'm going to recess. When he
gets like this, I get very nervous.
DR. KRESS: If I want to know what the spiking
factor is under a main steam line break accident, where I've
got leaky steam generator tubes, I have no idea what it is,
from what I heard. I have no idea, because we do not have
any data at all related to that subject.
I don't know whether it's 500 or 5,000 or five.
That's not a question. It's just a comment.
DR. POWERS: We'll take it at that and we can
puzzle it over the recess.
[Recess.]
DR. POWERS: Since you're not Joe Muscara, I
assume you must be Ken Karwoski. He doesn't look like Joe
Muscara. The floor is yours, sir.
MR. KARWOSKI: Thank you. Good morning. My name
is Ken Karwoski. Today, with the assistance of Joe Muscara,
I'd like to discuss our three issues with respect to steam
generator tube integrity.
The order in the package is a little different, as
Jack Strosnider indicated. The first issue I would like to
discuss is the regulatory framework and operating experience
to date. The second issue I would like to discuss is the
technical basis for Generic Letter 95-05, the voltage-based
repair criteria, including a discussion of the leak and
burst correlations.
And then the third issue I would like to discuss
are the capabilities and limitations of NDE with respect to
detection and sizing of flaws.
The guidance with respect to steam generator tube
integrity is located in various places. The general design
criteria, 10 CFR 50, Appendix A has general requirements
with respect to the integrity of the reactor coolant
pressure boundary. Appendix B deals with quality assurance
requirements. Part 100, which you've heard about this
morning from Jack Hayes, and dose limits.
Regulatory Guide 1.121 contains guidance with
respect to the loadings that the tubes should be able to
withstand. Regulatory Guide 1.83 discusses in-service
inspection guidance.
The standard review plan addresses various things,
such as in-service inspection; also, the design of the steam
generators and water chemistry, to some extent. The ASME
code has various repair criteria and the technical
specifications.
The plant technical specifications, as you heard
this morning, were developed about 25 years ago, when the
prevalent forms of degradation were general wall thinning.
The degradation that we're observing today was not
anticipated or tech specs typically specify a depth-based
tube repair criteria based on that general wall thinning
type of phenomenon.
The requirements for the inspection, repair, and
for normal operating primary to secondary leakage are
contained within the technical specifications.
The typical technical specifications in plants
today, plants that have not implemented an alternate repair
criteria, are listed on this slide. The first thing in the
technical specifications is the sampling program.
The basic sampling program involves a three
percent initial sampling of the steam generator tubes. That
sample is expanded based on the categorization of C-1, C-2
and C-3.
Basically, what those categories are, it says if I
have so many tubes or a certain percentage of tubes that are
either degraded or defective, I need to inspect more tubes.
A C-3 classification can result in 100 percent inspection.
Most of the plants with extensive degradation would end up a
C-3 classification.
The sampling program in the technical
specifications also require a reexamination of all
previously degraded tubes. With respect to the frequency of
inspection, the technical specifications simply say that
once every 12 to 24 calendar months, you should do an
inspection. That can be lengthened to 40 months based on
the categorization of C-1, C-2 or C-3, and it can be
shortened to 20 months.
It also requires inspections after certain events,
such as tube leaks in excess of the normal operating limit,
seismic occurrence, LOCA, and the steam line break.
With respect to the extent of the inspection, it
basically says that you are required to inspect the hot leg
of the tube around the U-bend to the top support plate on
the cold leg side.
The technique for inspection is not specified, and
I already mentioned that the repair criteria is typically 40
percent of the through-wall and it's applicable to all forms
of degradation.
So that's what you will find in most technical
specifications today for plants that haven't implemented an
alternate repair criteria.
The 40 percent depth-based limit was based on
guidance in Regulatory Guide 1.121. There are several
structural criteria that the tube is required to meet.
Typically, the most limiting is that the tube should be able
to withstand a pressure differential of three times the
normal operating pressure or 1.4 times an accident
differential pressure, and, typically, the most limiting is
the steam line break.
In addition, Regulatory Guide 1.121 indicates that
the normal operating primary to secondary leakage limit
should be based on a limiting crack length, that length that
would be limiting in terms of the structural criteria of
three delta P or 1.4 times steam line break.
Down here, basically what I have is a simple
derivation of the 40 percent plugging criteria. Basically,
assuming a general wall thinning, you need 40 percent of the
tube wall in order to withstand the three delta P or 1.4
times steam line break, and if you include an allowance of
ten percent for growth and ten percent for NDE uncertainty,
you would arrive at 40 percent repair criteria.
DR. POWERS: Let me understand this. The 40
percent is the result of considering NDE error and
uncertainty.
MR. KARWOSKI: Regulatory Guide 1.121 indicates
that both NDE uncertainty and crack growth need to be
accounted for in the plugging limit.
DR. POWERS: Right.
MR. KARWOSKI: Given that there is a -- the repair
limit is 40 percent, there is roughly a 20 percent margin
for both. Whether or not it was explicitly called out, the
ten percent, each one.
DR. POWERS: But the bottom line is you're saying
with 20 percent of the wall, you can meet the three times
normal operating or 1.4 times maximum allowable.
MR. KARWOSKI: With 40 percent of the wall. With
40 percent of the wall, you would be able to withstand
roughly three times --
DR. POWERS: I guess what I'm asking is if I had a
tube --
MR. KARWOSKI: It's 60 percent.
DR. POWERS: If I had a tube that I absolutely
knew had 20 percent of the wall there, an NBS standard tube,
if you will, had 20 percent of the wall left, would I be
able to meet -- now, I think we've had tests that said you
do with 20 percent.
MR. KARWOSKI: I don't believe so. It's roughly
40.
MR. STROSNIDER: I'd suggest it depends on the
type of degradation. For the analysis Ken's talking about,
where the tube was assumed to be uniformly thinned, you
would still meet ASME code, if you were uniformly thinned
and the amount of material missing was 60 percent.
So if you had 40 percent remaining, you'd still
meet the code allowables.
When you start looking at cracks and other types
of defects, you may be able to withstand something deeper
and still meet the code factors of safety.
DR. POWERS: Okay. I think I understand. Thank
you.
MR. STROSNIDER: It depends on the length of the
flaw.
MR. KARWOSKI: So what are some of the issues?
Over the last few days, you've probably identified a lot of
issues with the current regulatory framework. The major
goal of the steam generator tube inspections is to ensure
the structural and leakage integrity for the operating
interval between inspections.
Structural integrity per Reg Guide 1.121 and the
ASME code, and leakage integrity per Part 100 and GDC-19, as
was pointed out this morning.
As you know, the technical specifications do not
reflect either the current degradation modes or the
inspection technology that we have today. The repair
criteria of 40 percent tends to be conservative for cracks.
The inspection sample size, the expansion criteria
and frequency do not explicitly take into consideration the
severity of the degradation.
It is based on that classification of C-1, C-2 and
C-3, which more is a function of the number of tubes that
either exceed the repair limit or are degraded.
And as we know, the leakage limits don't prevent
tube burst.
As a result of these shortcomings, the NRC and the
industry have been taking action over the last several
years, for quite a long time. I've listed here various
efforts that have been underway. Some of the industry
efforts are that they have improved their examination
guidelines. I think the original version came out sometime
in the early to mid 1980s.
They have subsequently revised those several times
based on lessons learned and based on the changing forms of
degradation and the technology.
The industry has a steam generator management
program which actively participates with the NRC on various
steam generator issues.
In addition, the industry, as Jack Strosnider
pointed out this morning, in NEI-97-06, they've adopted a
condition monitoring and operational assessment philosophy.
I'll discuss those a little later, but basically what that
involves is condition monitoring. It is a backwards look to
make sure that you operated safely during a cycle.
Operational assessment is a forward look to make sure that
you can safely operate during the period of time before your
next inspection.
With respect to some of the NRC efforts, the NRC
has issued, over the last ten years, several generic
letters, Generic Letter 9503 on circumferential cracking of
steam generator tubes, Generic Letter 9705 on steam
generator tube inspection techniques, and Generic Letter
9706 on the degradation of steam generator internals.
We've also issued numerous information notices on
various topics, including sleeves, plugs, U-bend degradation
and other inspection related issues.
In addition, the NRC has an extensive research
program with respect to steam generator inspection and
repair criteria.
As I previously mentioned, NEI has their
guidelines, NEI-97-06. The staff has also developed draft
regulatory guide DG-1074, both of which address tube
integrity. Those are now being rolled up into an industry
initiative to address some of the issues with respect to
steam generators.
With respect to what we've observed to date, this
picture just shows some of the -- shows a lot of the
degradation mechanisms affecting steam generator tubes.
Just to go over some of the more prevalent ones,
in the tube sheet region, which is depicted in these
pictures, you have a variety of degradation mechanisms,
including the buildup of sludge on top of the tube sheet.
You have axial outside diameter stress corrosion cracking,
pitting and wastage can occur in the sludge pile.
At the expansion transition, the region of the
tube where it goes from expanded to unexpanded, you have
both circumferential and axial primary water stress
corrosion cracking and outside diameter stress corrosion
cracking.
At the tube support plate elevations, you can have
fretting, wear and corrosion thinning, the dominant
degradation mechanism back in the '70s. You can have axial
oriented outside diameter stress corrosion cracking and
intergranular attack.
At dented intersections, we've observed axial
primary water stress corrosion cracking. We have also
observed circumferential outside diameter stress corrosion
cracking and primary water stress corrosion cracking.
We've also observed fatigue at the upper most tube
support plates and also in the wedge reason of B&W plants,
although that picture won't show that. They have
once-through steam generators rather than U-tube steam
generators.
We've also observed free span cracking, free span
outside diameter stress corrosion cracking, and we've also
observed cracks in the U-bend.
DR. POWERS: In the U-bends, do you have -- I
don't know how to describe it well -- a bend is made and
it's too much and so they bend it back, so you get kind of
reverse bends on things.
MR. KARWOSKI: Sometimes that occurs, but --
DR. POWERS: And is that a site of --
MR. KARWOSKI: That has been a site of corrosion.
So what are some of the factors affecting tube degradation?
I think Dr. Hopenfeld yesterday touched on many of these.
Tube material, including the heat treatment. The
degradation mechanisms that I just had up there are
primarily observed in alloy-600 mill annealed steam
generator tubing. That's basically most of your older
plants, with their original steam generators.
There has been relatively little degradation in
alloy-600 thermally treated steam generators, which are the
later vintage of steam generators and some of the initial
replacement steam generators.
The tube material of choice these days for the
replacement steam generators are alloy-690 thermally
treated.
DR. POWERS: Now, I understand that in Europe,
they use something else, 800 alloy maybe.
MR. KARWOSKI: In Germany, they have used
alloy-800.
DR. POWERS: And is there anything substantially
superior or inferior to 800 relative to 690 and 600?
MR. KARWOSKI: I don't think I can --
MR. BALLINGER: The 800 works a little bit better
if you're in phosphate chemistry and the like. It doesn't
waste, doesn't get wastage like 600 did, does. So
replacement steam generators in Europe are all going to be
pretty much 690. I don't think there's any 800 that's going
to be used for replacement generators.
MR. KARWOSKI: Other factors affecting tube
degradation, grain size, carbide distribution, the
fabrication of the tubes and stresses. For example, the
expansion joints at the -- where the tube goes from the
expanded to unexpanded region, there's been various means
for expanding those.
Initially, utilities would role expand those.
They tried to lessen the stresses at the transition. They
then went to an explosive transition and currently most
people now do a full depth hydraulic expansion.
DR. POWERS: Let me ask a question about carbide
distribution. I see in the many documents we've been
provided three types of carbide distribution, called,
imaginatively, one, two and three.
I think I understand one. I don't understand the
distinctions between two and three.
MR. KARWOSKI: I'm not sure of the reports you're
referring to. Other people may be more qualified to address
that later, too.
DR. POWERS: Okay.
MR. KARWOSKI: Tube support plate design and
material effects, tube degradation, operating temperature
and stresses, the water chemistry, operating time and the
presence or absence or crevices.
There have been a number of tube ruptures. I've
listed ten tube ruptures here. Some people call an event or
a leak at Fort Calhoun a rupture, I did not include that.
The definition of rupture I used here is leakage
in excess of the normal makeup capacity of the plant.
Just going over each one of these ruptures, as you
can see, there's ten listed. Of these ten, eight have
occurred in the U.S., two in foreign PWRs.
The first rupture occurred in 1975 at Point Beach.
The rupture was primarily attributed to wastage. The tube
wasn't pulled for destructive examination, but they think
stress corrosion cracking may have also played a role.
The next steam generator occurred at Surry-2 in
1976. That was axial primary water stress corrosion
cracking up in the U-bend of the steam generator.
The Surry rupture was attributed primarily to
denting at the uppermost -- at the tube support plates
forcing the tight radius U-bend tubes closer together and
parting a stress up in the apex of the U-bend.
The Doel rupture occurred in 1979. It was also
axial primary water stress corrosion cracking in the U-bend.
However, in that case, they attributed the rupture to the
bending process and the fact that there was ovalization of
the tube that wasn't in accordance with specifications.
In 1979, there was a rupture at Prairie Island due
to a foreign object. In Ginna, in '82, another foreign
object. There was some discussion on the magnitude of the
leak rates from a rupture. This rupture was on the order of
760 gallons per minute.
In 1987, there was a rupture at North Anna which
was attributed to fatigue at the upper most tube support
plate. Some of the factors affecting that was denting and
improper or inadequate ABB support up in the U-bend region.
In McGuire, in '89, there was a free span rupture
was a result of axial outside diameter stress corrosion
cracking. That was on the cold leg side. The crack was
associated with a manufacturing scratch.
In 1991, there was a rupture at Mihama, which was
very similar to the event at North Anna-1.
In 1993, at Palo Verde-2, there was a rupture as a
result of free-span cracking. This was on the hot leg side.
The utility attributed that, in part, to a dryout region,
which they refer to as an arc, which is present in the outer
periphery of the bundle up in the top. Then Indian Point-2
in February of this year, which was a result of primary
water stress corrosion cracking in the U-bend.
In addition to ruptures, there have been a number
of leaks that have resulted in forced outages. Basically,
what I show here is the number of forced outages as a
function of year.
As you can see, back in the '70s and early '80s,
there is a number of forced outages for a variety of
reasons. Here in the '90s, there have been -- if you look
at this literally, you could say that there has been a
decreasing trend in the number of forced outages as a result
of leakage.
Some of the shutdowns that are on that graph were
initiated because the plant exceeded the primary to
secondary leakage limit in the technical specification. In
the standard technical specification, that limit is
typically around 500 gallons per day through any one steam
generator.
Some of these plants shut down voluntarily before
the leakage exceeded those limits. In addition, I wanted to
point out that some plants have operated with leakage over
the course of a cycle and then shut down at the standard
refueling outage.
Some of the causes of shutdowns as a result of
leakage in the 1990s include sleeves, primarily the B&W
kinetically expanded sleeves. That was the result -- that
was the cause of the Trojan leak back in the '92 timeframe.
MR. HIGGINS: The typical tech spec limit that you
mentioned of 500 gallons per day, the new NEI document that
Jack had mentioned that the plants had all committed to has
got a number of 150 GPD. Does that mean that the plants are
now all observing that versus the 500?
MR. STROSNIDER: The plants have implemented
administrative limits reflecting the 97-06 guidelines.
MR. HIGGINS: Thank you.
MR. KARWOSKI: So that the B&W kinetically
expanded sleeves resulted in a number of leakers back in the
early '90s. The last steam generator that has these sleeves
installed is being replaced now at ANO-2.
There's been leaker outages in the '90s as a
result of plug leakage, loose parts, fatigue primarily in
the B&W once-through steam generators, in the lane wedge
region or in the area bordering the lane wedge region, and
leakage has been observed as a result or forced shutdowns
have resulted as a result of leakage due to stress corrosion
cracking at expansion transition, tube supports and free
spans.
A number of plants have replaced their steam
generators. I believe there's 25 plants that either have
replaced or are replacing. Replacements started in the
1980s. Cook finished in July of 2000. ANO-2 and Indian
Point-2 are currently replacing right now.
With respect to the tube materials, I'll just
point out that these early replacements were all alloy-600
thermally treated. When you got to Cook-2, most of the ones
after this are alloy-690 thermally treated, with some
exceptions.
Palisades used steam generators available at
another plant, so I believe these are 600 mill annealed.
Salem also used previously available steam generators at a
cancelled plant, so these are 600 thermally treated. And
the Indian Point-2 steam generators, I believe, are 600
thermally treated. The rest are 690.
A number of plants also have indicated that they
plan on replacing steam generators. Some of these
replacements are a result of tube degradation or as a -- or
in combination with license renewal, they believe that they
will need new steam generators in order to operate 60 years.
To date, only steam generators from Westinghouse
and CE have been replaced. Some of the replacements in the
next ten years will probably be in B&W units, as well.
As a result of all the tube degradation, a number
of utilities have proposed various alternate repair
criteria. One of the first alternate repair criterias was
for degradation within the tube sheet region.
When you have a tube fully expanded against the
tube sheet, you only need a certain length of engagement in
order to ensure that the tube does not pull out during
accidents. The remainder of the tube can be degraded
without any significant effect on the structural or leakage
integrity, and that's what these repair criteria are
basically for, degradation in the tube sheet area.
Of more interest are the alternate tube repair
criteria that has been implemented at the tube support plate
elevation with respect to predominantly axially oriented
outside diameter stress corrosion cracking.
I've listed the plants that currently have this
implemented. There have been other plants, but they have
either subsequently replaced their steam generators or
ceased operation.
Beaver Valley, Comanche Peak, Diablo, Farley, Kuwanee,
Prairie Island, Sequoyah and South Texas currently have
repair criteria and, I believe, as Jack Hayes indicated,
there's others that are being reviewed now.
There's also been an alternate tube repair
criteria for axially oriented primary water stress corrosion
cracking at or near dented tube support plates. That's been
approved on an interim basis at Sequoyah.
That concludes the first part of the presentation
with respect to the regulatory framework and operating
experience. I'd be happy to answer any questions on that.
The next part of the presentation deals with
Generic Letter 95-05. This will tend to be lengthy. I'm
not sure if you -- do you want to start this at this point?
DR. POWERS: Why don't we go ahead and start it.
MR. KARWOSKI: Okay.
DR. POWERS: And I presume that there will be some
point in there that it's logical to take a break. Or is it
continuous?
MR. KARWOSKI: I think this one might be
continuous.
DR. POWERS: Okay. I run into problems starting
early and starting late is okay, starting early is a
problem. So I think what we will do is just interrupt you
at 12:00.
MR. KARWOSKI: Okay. Generic Letter 95-05
addresses one form of degradation, axially -- predominantly
axially oriented outside diameter stress corrosion cracking
at the tube support plate elevations.
There's two fundamental goals of the repair
criteria in this generic letter, to ensure adequate
structural and leakage integrity.
The evaluation of structural and leakage integrity
require periodic inspections. It requires correlating those
inspection parameters with the tube structural and leakage
integrity and evaluation of the tubes accepted for continued
service.
And I apologize, this is where I've jumped ahead
in the presentation. It's page 10-27. And I also
apologize, as a result of some of the presentations
yesterday, there are some additional slides that I have
prepared to address some specific comments.
DR. POWERS: Thank you.
MR. STROSNIDER: Ken, I think we might also
mention, I think we provided -- there was discussion
yesterday of some of the proprietary data and I think we
provided copies of that information for the panel, or we
will.
MR. KARWOSKI: Yes. This is the proprietary
information containing the data in the database.
MR. STROSNIDER: And I would just point that
because of it's proprietary nature, we won't be presenting
it on the screen, but the members of the panel will have it
so they can look at it, and, of course, need to treat it as
proprietary.
MR. KARWOSKI: Okay. So just so that everybody
understands what degradation mechanism we're talking about,
Generic Letter 95-05 primarily addresses axially oriented
outside diameter stress corrosion cracking at the tube
support plate elevations.
It does not permit -- or it does not apply to
circumferential cracks, primary water stress corrosion
cracks or cracks that go outside the tube support plate, and
it does not apply to general wastage or thinning.
As I mentioned, in order to ensure the structural
and leakage integrity of the tubes, you need to do
inspections. The generic letter specifies specific
inspections that must be performed. Take this into context
of the current regulatory framework, which says three
percent initial inspection and expand based on the results.
GL-95-05 requires the licensees to perform 100
percent bobbin coil inspection; basically, all the way
around to the cold leg, to the point where they've observed
degradation, and a 20 percent sample at the next tube
support plate elevation.
I say that because that's what is in the generic
letter, practically speaking, everyone does 100 percent tube
end to tube end.
The bobbin coil allows for a rapid screening of
the tubes for defects. The extent of the degradation is
measured in terms of the voltage response for the defects at
the tube support plates.
There are detailed procedures to ensure that the
voltage response of the degradations being measured in the
field is comparable to those in the structural and leakage
integrity databases.
So basically it tells the analyst what size probe
to use, what frequency mix to use to size the degradation.
It instructs them to record the maximum voltage response at
that location, as the voltage.
So there are detailed procedures with respect to
the data analysis.
MR. HIGGINS: Can you say what that database is?
MR. KARWOSKI: I'll get into the database in a few
slides. In addition to the bobbin coil examinations,
rotating pancake coil examinations are performed. This
permits a better characterization of the defects to ensure
that degradation is confined within the tube support plate
and is predominantly axial.
Make sure that you're not applying this to
circumferential degradation or other forms of degradation;
that you can get some idea of the morphology as a result of
the rotating pancake coil.
DR. POWERS: You tend to speak of cracks as axial
or circumferential. Is there a case where things are at an
angle?
MR. KARWOSKI: Absolutely.
DR. POWERS: And how do you make a distinction
between axial and circumferential when you're at an angle?
MR. KARWOSKI: There is some analyst judgment
involved. With respect to that specific issue, if you look
at the database, clearly, from the metalography, you can
tell -- you will see that the degradation at the support
plates occurs in networks or is a cellular type of
corrosion.
So there are oblique angles. There may be short
segments that are circumferential in extent. With respect
to the eddy current data evaluation, though, those
typically, what you will see is you will see a pattern.
Usually those short circumferential extents will
not be discerned in the NDE examination.
MR. SIEBER: So the NDE doesn't tell us about
circumferential cracks.
MR. KARWOSKI: It cannot readily detect the short
segments. It will find, and that will be this afternoon, it
will find distinct circumferential cracks or large
circumferential components. It can do that, and I will
present some data to show that.
MR. STROSNIDER: Ken, and you may get to this
later, it might also be a time to interject that there is a
tube pull requirement associated with it. Are you going to
talk about that?
MR. KARWOSKI: Yes.
MR. STROSNIDER: Okay. But I would just interject
that plants are required to periodically pull tubes to
verify that the degradation mechanism is consistent with
what's in the database and what they've seen in the past,
but Ken will talk about that.
MR. KARWOSKI: These rotating pancake coil
examinations are performed at intersections with degradation
exceeding specific voltage limits and I will discuss the
limits, but basically it's one volt for three-quarter inch
tubing and two volts for seven-eighths inch tubing. There
are two correlations, depending on the size of the tubing.
The plants -- the Westinghouse plants that this
affects are either three-quarter or seven-eighths inch
tubing.
DR. CATTON: And you find these with the bobbin
coil.
MR. KARWOSKI: You find these indications with the
bobbin coil.
DR. CATTON: Then you do a detailed evaluation
with the rotating pancake coil.
MR. KARWOSKI: That's correct, and I will discuss
a little more on what you do with the rotating pancake coil
examination results.
You also perform these rotating pancake coil
examinations at tube support plate elevations, where the
dents exceed five volts. Part of the reason for doing that
is because in highly dented intersections, the bobbin coil
is relatively ineffective. The rotating pancake coil gives
you a better inspection.
You also perform these examinations at tube
support plate elevations with copper deposits. The reason
for that is because the pancake coil will give you a better
inspection. And also at locations with large mixed
residuals, for the same reason.
DR. POWERS: Maybe you should explain what you
mean by mixed residuals.
MR. KARWOSKI: What mixed residuals are are when
you do these inspections at the tube support plate,
depending on the frequency, you will get a response not only
from the tube, but also from the support plate.
So what you do, you're using a multiple frequency
probe, you will -- you take one frequency that is more
sensitive to the tube and another frequency that is more
sensitive to further out or the tube support plate and you
essentially mix out the signals. That's not a 100 percent
perfect. There are some what's called residuals and so if
you have large mixed residuals, you will inspect those with
the rotating pancake coil to give you a better examination.
MR. STROSNIDER: Ken, I guess that large mixed
residual, it basically looks like distortion of the eddy
current signal.
MR. KARWOSKI: Yes.
MR. STROSNIDER: When the analyst looks at it,
it's an amount of distortion that's in the signal, because
the mixing isn't perfect.
MR. KARWOSKI: So I've discussed the inspections
and I've tried to give you an idea that there are detailed
procedures in order to interpret the voltage and to
characterize the defects, but that's only one part of the
picture.
You also have to have correlations correlating
that inspection parameter to the structural and leakage
integrity of the tubing.
The correlations come from two primary sources,
tubes removed from operating steam generators and specimens
produced in model boiler facilities. The specimens produced
in model boiler facilities span a larger range than the data
from tubes removed from operating steam generators, because,
in general, the voltages observed in the field typically
aren't as great as you can produce in a model boiler.
And as Jack Strosnider pointed out, there is a
periodic tube pull program for confirming the degradation
mode at the plant. There's an initial tube pull that
involves a couple tubes and, I believe, four intersections.
After that initial tube pull, there is a periodic
tube pull requirement, which involves pulling additional
intersections at a frequency of about every two or three
outages.
The examinations performed on these tubes, I'll
start down here. You perform a metallurgical examination to
make sure that the degradation mechanism is consistent with
that observed at other plants and for -- and is consistent
with the other data in the database.
You also do leak testing, what's involved here is
the tube is pressurized internally and on the outside. It
is taken up to steam line break pressure at temperature of
around 600, 650 degrees Fahrenheit.
The first thing that they do is determine whether
or not it leaks or not. If the tube leaks, then they go on
to measure the leakage. If the tube doesn't leak, then they
just use the data in the probability of leakage correlation,
which I will be discussing later.
DR. POWERS: In all cases, the testing is done at
temperature?
MR. KARWOSKI: I don't know if I can say all
cases, because I don't recall from that --
DR. POWERS: We'll accept 90 percent.
MR. KARWOSKI: The vast majority, and I would also
like to point out, but I need to point out, it may not be
the exact temperature and there may need to be some
adjustments to the data. It may be taken at 2,603 PSI
instead of 2,650 and maybe taken at 580 degrees F instead of
620. So there are adjustments that need to be made to the
data.
The burst testing is performed at room
temperature. Typically, after the leak testing, they will
take that specimen, they will insert a bladder. The reason
for the bladder is to prevent excessive leakage from
preventing them to achieving burst.
They will burst test the tube at room temperature and they
use that burst pressure in the correlations. When I get to
the slides on the burst pressure correlation, all the data
is normalized to a specific burst pressure and they use
lower tolerance. Then they scale it up to operating
temperature and take a lower bound, and I'll point that out
on the correlations.
So that's the testing that is performed. Earlier,
I discussed the regulatory criteria. Typically, the most
limiting is that the tubes must withstand a pressure
differential of three times the normal operating or 1.4
times the maximum postulated accident or steam line break.
This roughly turns out to be around 3,660 PSI.
For degradation at the support plate, during normal
operation, the plate is present. I should point to this one
because this is the degradation mode. That plate is
present.
As a result, the criteria for three times the
normal operating pressure is met during normal operation.
The degradation is confined to within the tube support plate
region.
Given the clearances between the tube and the tube
support plate, that tube will not burst. So the three delta
--
DR. POWERS: I guess the question comes up, when
you say it's combined within the support plate, what exactly
does confined mean?
MR. KARWOSKI: The degradation, in general, does
not exceed -- does not extend above or below the tube
support plate.
DR. POWERS: At all.
MR. KARWOSKI: There have been pulled tube data
where there has been some minor extension of the outside
diameter stress corrosion cracking beyond the plate. To my
knowledge, that has only been discovered as a result of
destructive examination in the cases I'm familiar with and
it's only on the order of .02, .03 inches beyond the plate,
and it's typically attributed to some slight deposits which
basically come up along the side of the tube.
DR. KRESS: If that is found by the NDE techniques
to extend beyond, then that's excluded from this being
confined.
MR. KARWOSKI: There is a reporting requirement in
the tech specs that if they find that degradation, they have
to let us know because it will draw on the question, the
validity of all the arguments.
If you look at our understanding of this
phenomenon, it's basically crevice corrosion in this
location, all the pulled tube data has suggested that
degradation is confined within that support plate region.
If they find it by NDE, there is a reporting requirement to
address that.
DR. KRESS: Then it's treated like a crack that's
outside the confined area.
MR. KARWOSKI: Yes. We would have to question
whether or not they should even implement the repair
criteria not only at that location, but at other locations
in the plant. That's the purpose of the reporting
requirement.
DR. BONACA: But can you detect it if you have
just a fraction of an inch?
MR. KARWOSKI: That pulled tube data that I was
referring to where there was a minor extension, that was not
detected in the field. A .02, .03 inches will not be
detected in the field.
On the other hand, it probably will not have a
significant effect on the burst pressure of that tube.
MR. BALLINGER: A question. Circ cracks are
plugged on detection, right?
MR. KARWOSKI: Yes.
MR. BALLINGER: Getting back to this degree of
circumferentiality issue, when you have a network of cracks
in the TSP, in the support plate region, is there some kind
of judgment that has to be applied to -- since you can't see
that with a bobbin coil and the rotating pancake doesn't
work too well either for that kind of situation, what
happens if there is a likelihood that you've gotten an
equivalent circumferential crack there?
How do you deal with that?
MR. KARWOSKI: That will show up in the burst
pressure database. The correlations are all empirical. So
you've pulled a variety of tubes with given voltages. You
also have a number of tubes produced in model boiler
specimens.
Those tubes are somewhat representative of what's
out in the field or they are representative of what's out in
the field. During those tests, if you were to have a,
quote-unquote, limiting circumferential crack, you would
have observed a circumferential failure.
That's not what we've been observing today.
That's part of the reasons for the periodic tube pull
examinations to confirm that that is not occurring.
MR. BALLINGER: So that would be picked up in the
tube pull.
MR. KARWOSKI: Tube pull and if you do a -- when
you do your rotating pancake coil examinations, which are
basically of most indications above one or two volts, if you
had a large circumferential extent, you would notice.
DR. CATTON: When you do a tube pull, do you
literally snake that entire tube out of there?
MR. KARWOSKI: Basically, what's done is they will
cut the tube at a specific location.
DR. CATTON: And pull.
MR. KARWOSKI: And pull it through the tube sheet.
DR. CATTON: What do they do with the rest of the
tube, it just stays there?
MR. KARWOSKI: It just stays there. They
frequently stabilize it, depending on what they believe the
tube will whip around, if they believe there's going to be
some damage, but they'll stabilize that tube.
MR. STROSNIDER: Ken, we might mention, too, I
think there was a little discussion yesterday. During the
tube pulling process, and it's not always easy to pull these
tubes out. There can be some --
DR. POWERS: Is it ever easy to pull these tubes
out?
MR. STROSNIDER: There's a possibility of some
change in terms of the defect that you're trying to get at
and, again, I don't want to get ahead of you, Ken, but I
think when you look at what's plotted in the database, it's
the in situ voltage versus what was tested after you pulled
it. And during the pulling it, it's possible that ligaments
might tear or whatever, but in general, the pulling is not
going to make the burst pressures or leakage characteristics
better. It's going to make it worse.
So there is some conservatism in that.
MR. KARWOSKI: Right. And if you did have a large
circumferential network which was limiting, when you're
pulling that tube, there's some extreme forces, it would
break.
And that has occurred for some circumferential
cracking at the top of the tube sheet where the licensees
have attempted to get those specimens out. They go to the
tube pull, they pull it and basically it rips. But that has
not been observed at the support plate elevations for which
this generic letter is applied.
As I discussed, there are two correlations. This
doesn't have any data, it's old, so it won't match the data
that I've presented or that I've provided to you, the
proprietary information, but it will help me illustrate the
points.
This correlation is for seven-eighths inch
diameter tubing. We have the burst pressure over here and
we have the bobbin voltage on a log scale over here.
If you look in your package, you will see the data
point scattered throughout. You will see a mean regression
curve, a lower 95 -- a mean regression curve where the data
has been normalized to specific material properties, a lower
95 percent prediction interval.
DR. POWERS: It's the 95 percent confidence level
for a prediction drawn from the correlation?
MR. KARWOSKI: This is the 95 percent prediction
interval associated with this mean regression curve.
DR. POWERS: What I find remarkable about that
curve is that as you move away from the mean of the data,
those curves typically expand out a lot. In principal, they
go to infinity at -- well, they go to zero and on the other
side they go to infinity, if you get far enough away from
the mean of the data, and this curve does not seem to do
that.
MR. KARWOSKI: You asked me that question a number
of years ago and I did research after. If you were to blow
this up and expand the scales, you would see the exact
effect that you're talking about.
You just don't notice it on the scales here, but
you are absolutely correct. When you blow that up, you see
that -- see the curves with that trend, the blow-up way down
here.
DR. POWERS: I'm confident that -- I mean, I'm
encouraged that my intuition is good. I'm surprised I don't
see it, because it did look awfully scattered.
MR. KARWOSKI: Yes. But you do observe it when
you blow this up.
As I mentioned, this is the mean curve adjusted to
a specific set of material properties. The lower 95 percent
prediction interval. Because material properties in the
steam generator tubes vary, they adjust that for the lower
95 percent material properties and they get this dotted
curve down here.
In order to determine the repair limit, basically,
you take the intersection of this curve with your limiting
regulatory guide pressure, which is around the 3660 PSI, you
come down, you get a limit of 8.8 volts.
That is then consistent with Regulatory Guide
1.121. You take off allowances for growth and NDE
uncertainty, and I'll discuss this in a little bit, and you
get a repair limit at which tubes would need to be plugged,
and I will discuss that because this is not what we've
accepted as a result of some of the assumptions made in the
growth and NDE uncertainty.
This next viewgraph here, I'll just discuss it
from this. The industry's original proposal said basically
we would like to implement a five and a half volts.
Anything above five and a half volts we'll leave in service.
I'm sorry. Anything less than five and a half volts we'll
leave in service, anything greater than, we would plug.
The values have changed and it's evolved over
time, but basically the staff was concerned with this
approach given that back in the '95 timeframe, most of the
data out here was from model boiler specimens. The pulled
tube data was relatively scarce and it was all centered in
the lower voltage regions.
Because of that, because you can have higher than
average growth rates which are used in the calculations and
your NDE uncertainties aren't limited and a variety of other
reasons, the staff chose to use lower voltage limits.
In the case of seven-eighths inch diameter tubing,
we chose two volts, and, for three-quarter inch tubing, one
volt because of differences in the correlation.
MR. SIEBER: Just a quick question. The 1.4 times
the steam line break differential, the other requirement is
three times the normal operating differential, which I
presume, for Westinghouse steam generators, is either 1550
or 1600, depending on the model.
So that would come out to be 4800 on that chart.
MR. KARWOSKI: That's right. But if you remember
from this plot here, during normal operation, this plate
will be in place.
MR. SIEBER: Okay.
MR. KARWOSKI: That plate is in place. That three
delta P, that tube is not going to --
MR. SIEBER: So you don't consider it.
MR. KARWOSKI: Right.
MR. SIEBER: You don't consider that, okay.
MR. KARWOSKI: We don't consider it. So how are
these repair limits implemented? This is not in your
handout, I don't believe. This is a subsequent -- I didn't
plan on getting into all of this.
Below the lower bobbin voltage repair limit, and
what I mean by that is if you go in and inspect and find
something less than one or two volts, you can allow those
tubes to remain in service.
Those are tubes that can be left in service.
DR. POWERS: Yesterday we had several mentions of
three volts.
MR. KARWOSKI: I will get into that at the very
end, but I want to point out that that three volt criteria,
although it is a modification of this approach, it is not
the same. It is not the same as what's in Generic Letter
95-05.
There are similarities and some of the data is the
same, but there are differences.
MR. HIGGINS: On the last slide, you said that you
had implemented a lower repair limit as opposed to the five
and a half.
MR. KARWOSKI: Right.
MR. HIGGINS: And you're saying now that it's one
or two.
MR. KARWOSKI: It's one volt for three-quarter
inch diameter tubing and two volts for seven-eighths inch
diameter tubing.
Between the lower voltage repair limit, this one
and two volts, and the upper bobbin voltage repair limit,
which would be the equivalent of the 8.8 volts that I showed
you -- I'm sorry -- the 5.5 volts, you need to do RPC
inspections for all those indications. That confirms your
degradation morphology and will give you added confidence
that the degradation is within the support plate.
Any indications that are not confirmed by RPC can
remain in-service. What I mean by not confirmed is the
bottom coil has a certain detection threshold. The RPC has
a certain detection threshold. The bobbin tends to be more
influenced by noise and other masking features. You might
call something that is not a flaw a flaw, or the RPC's
threshold of detection is different.
If you don't confirm it by RPC, then you can leave
that tube in service, and the reason is that the RPC is
typically less sensitive to interfering signals. So things
that you might have caught with the bobbin may not actually
be flaws.
However, they may be flaws, but RPC is less
sensitive to shallow crack networks, but it's at least
equally sensitive to deep cracks.
So what that means is even though that the RPC
isn't seeing it, it's probably not significant.
DR. POWERS: Now, that presumes that shallow crack
networks are not going to coalesce and make a deep crack.
MR. KARWOSKI: That will be handled in the growth
rate analysis that I'll get into in a minute, but yes and
no. I will point out that even though these tubes are
allowed to remain in service, the industry, I believe,
originally argued that they should not include them in the
probabilistic calculations that I'll be discussing.
The staff said yes, you need to include those,
even though you don't believe there is degradation there.
DR. POWERS: Okay.
MR. KARWOSKI: You need to address them.
DR. POWERS: You're setting the stage for that
part of your talk that will take place after the lunch
break.
MR. KARWOSKI: Actually, this might be a very good
ending point. Above the upper bobbin voltage repair limit,
the indication must be repaired regardless of RPC results.
Licensees are required to RPC even those tubes that they are
going to need a plug, for the obvious reason. Those are the
ones that will start probably showing circumferential
extent, extending outside the support plate.
We wanted to make sure that there is nothing going
on there that we need to be aware of before the repair
criteria is implemented.
DR. CATTON: And repair means plug.
MR. KARWOSKI: Plug or sleeve. If the plant is
licensed to sleeve.
DR. POWERS: And the extent of Ken's talk here is
a prologue for all this in-depth analysis he's going to do
for us. Are there any questions you want to pose to him
now?
MR. STROSNIDER: If I could interrupt for just a
second, Ken. I don't know if this is the best time or not,
I'm going to put him on the spot here. With regard to the
correlation in the database for burst pressure, I just
wanted to point out that, I think you've got it in front of
you, there's a substantial number of data points and part of
the discussion about whether there -- first of all, it is an
empirical model. So some of the things you need to look at
are what sort of correlation coefficients, how much data,
can you do reasonable statistics with this.
And I would suggest, if you look at the amount of
data and the care that's been taken to make sure that it's
the right population to compare to the steam generators,
there is a good basis for doing this sort of empirical
evaluation.
And you'll see more when we get into the leakage
and other correlations.
MR. KARWOSKI: In the proprietary handout I gave
you, you have all the data or you have all the correlations
and some of the correlation coefficients that Jack was
talking about, but basically here is a summary for the burst
correlation, since we were discussing it, for three-quarter
inch tubes and seven-eighths inch tubes.
Basically, there's 96 and 91 data points,
correlation coefficient --
DR. POWERS: If I were to ask you a question, what
is the probability that a random data set would produce such
a high R-square given that there are 96 points, what would
you answer?
MR. KARWOSKI: I would probably refer to the
statistician.
DR. POWERS: R-squared values are virtually
useless. The important point to understand is what's the
probability the random data set would produce such a high
value of R-squared.
MR. KARWOSKI: Just since we're on that point, we
have had a statistician look at any of these correlations.
The statistician has had a tremendous impact on the leakage
analysis. We've had these correlations looked at.
DR. POWERS: I'm fascinated in what you have to
say about something at the 12 percent R-squared.
MR. KARWOSKI: That's the interesting one,
actually.
DR. POWERS: At this point, I want to recess, and
apologize to Ken for interrupting his presentation, it's
going awfully well, a very nice presentation, but let's
recess and come back at 1:00. At that time, Dr. Kress will
be chairing the session.
[Whereupon, at 12:05 p.m., the meeting was
recessed, to reconvene this same day at 1:00 p.m.]
. AFTERNOON SESSION
[1:05 p.m.]
DR. KRESS [Presiding]: Can we come back to order,
please? I will be the substitute Chairman till Dana gets
back, which may be not too long from now.
So we'll continue now where we left off before
lunch.
MR. KARWOSKI: Thank you. Okay, just to go over
quickly what I did this morning, this morning I showed the
first pressure correlation.
I showed you how we determine the tube repair
limits with respect to the one-volt, two-volts, and the
upper repair limit. I also discussed the inspections that
were performed.
That's the deterministic approach with respect to
addressing structural integrity, but as I pointed out, there
are several assumptions with respect to lower tolerance
limits, material properties using a lower, 95-percent
prediction interval.
As a result, from a structural integrity
standpoint for the entire steam generator, you need to take
a bigger look at what the cumulative effect of all the
uncertainties and of some of your assumptions with respect
to using an average growth rate or a 95-percent confidence
interval.
So, to ensure structural leakage integrity, the
Generic Letter requires two calculations: The first
calculation is a conditional probability of burst under
steam line break conditions.
And that calculation is necessary to ensure that
the repairs are adequate from a structural integrity
standpoint. And as I've just mentioned, it's because the
values of the growth in NDE uncertainty may exceed those in
the deterministic determination of the repair limits, it's
because we used the lower 95-percent prediction interval for
the burst pressure correlation.
It's also because we only used lower 95-percent
material properties. You can have values less than those.
It's also because you're only looking at a single
indication and not the entire steam generator. There is a
cumulative effect.
The other calculation that we do -- and I'll be
talking more about how we do that, in a minute -- is, we
determine what leakage will exist under postulated accident
conditions.
And that's necessary because with this
voltage-based approach, there is no correlation between
voltage and depth. As a result there are through-wall
cracks or near-through-wall cracks, can either remain in
service after the inspection, or they may develop --
service.
And as a result, we have to do a calculation to
determine the leakage under those conditions.
Both of those calculations will require some
knowledge of what's going to be present at the time of the
next inspection, because from a conservative standpoint, you
want to know, if I had that steam leak break at the end of
the next operating interval, what would my probability of
burst be and what would be my leakage?
That's going to be the most limiting, because the
tubes will have the chance to progress the most.
Looking at this big picture, basically what you do
is, you do your eddy current inspection, you find all the
defects at the tube support plates, because under a
100-percent inspection, you do your RPC examinations.
With all the indications you detect, you then make
a POD adjustment. We use the constant value of .6 as the
POD.
And I'll discuss the basis for the .6 later today
in then presentation on NDE capabilities.
DR. CATTON: What is POD?
MR. KARWOSKI: Probability of detection.
DR. BALLINGER: Is that the 40-percent level,
40-percent through-wall?
MR. KARWOSKI: We applied that .6 value to
everything that you find. Remember, what you're going to
have is at the beginning of the cycle -- this graph doesn't
really show it.
What you're going to have, after you do your
inspection, you're going to have a distribution of
indications. Some of those will be RPC-confirmed; some of
them won't.
They will be a function of voltage, okay? So it
has nothing -- any given voltage can have a range of depths
associated with it; there's no correlation.
We say that our probability of detecting a large
voltage indication is the same as our probability of
detecting a low voltage indication.
And we assume a constant value of .6 throughout
that range. So, if you were to say the worst degradation is
the largest voltage, and that the lower voltage indications
are minor, we're saying that you have an equal probability
of detecting both of those.
It's a very conservative assumption, which we'll
get into this afternoon when I talk about the POD.
In general, you would expect to find some of the
more significant flaws easier than some of the more minor
flaws.
So we applied this .6 value. So everything that
you find during an inspection, you divide it by .6, which is
roughly equivalent to multiplying it by 1.6-something.
And then you subtract out the indications that you
repaired. That will be your beginning-of-cycle
distribution.
It's very conservative in the sense that if you
found a ten-volt indication, you would assume that roughly
-- that you have roughly .67 of those indications left in
service.
DR. BONACA: Could you had a defect for which you
have no indication whatsoever?
MR. KARWOSKI: Yes, where you missed it during the
inspection.
DR. BONACA: So for this, really, the conservatism
doesn't apply, the statement of conservatisms doesn't apply,
because simply you don't detect it.
MR. KARWOSKI: But that's one of the purposes of
the .6 adjustment, is to account for that. It's to account
for the fact that you can miss flaws during your inspection;
that's one of the purposes. There is also another purpose,
and I'll get into that.
So you divide what you find during your inspection
by .6, and then you subtract off the indications that you
repaired.
DR. KRESS: Now, suppose you found no indications?
MR. KARWOSKI: If you found no indications, you
won't be implementing the repair criteria, you would just --
DR. KRESS: You don't assume that you might not
have detected --
MR. KARWOSKI: Utilities would normally not apply
for this repair criterion, unless they had a high likelihood
of finding it.
So, then you take this beginning-of-cycle
distribution, and you need to add in growth and NDE
uncertainty to get to the end-of-cycle distribution.
Then with that end-of-cycle distribution, that's
what you use to calculate your probability of burst.
DR. KRESS: Voltage growth is based on
extrapolating fast growth rate?
MR. KARWOSKI: The way voltage growth is
determined as part of this generic letter, is, when you go
in and inspect -- if today was your inspection, you would
inspect you'd find a distribution of indications. You would
have the voltages associated with those indications.
You then go back at that location, at your prior
inspection, and see what it was then, and you take the
difference, okay?
DR. KRESS: Fine.
MR. KARWOSKI: And the time, and you make the
appropriate adjustments for whatever your operating cycle
is, but you take the difference between those two, you
adjust it for the appropriate time, and that will be your
growth rate.
There are provisions in the Generic Letter that
address how many datapoints you need in order to use the
growth rate, the concern being is that if you didn't have
enough data, how could you project?
If this is one of your first cycles, you only have
40 datapoints, how well did you know that the growth rate of
the tubes?
And in those cases where there is limited data, it
requires the utility to use a bounding growth rate based on
other steam generators that are operated under a similar
condition.
DR. KRESS: And the assumption is that if the
growth rate is changing, that that change may not be enough
to worry about over one cycle?
MR. KARWOSKI: Yes, the assumption is that the
growth rates -- you're looking at this from a population
standpoint. The assumption is that the growth rates that
you observe during the course of that cycle will be -- when
you operate it the next cycle, adjusted for the appropriate
operating length, the growth rates will be comparable for
the entire -- for the population. That is an assumption.
The other thing that we do with respect to the
growth rate distribution, when you go in and do these
inspections and you find, say, a 1.5 volt indication, or --
that's probably not a good example.
Say you'd found a half a volt indication; you
could go back to the prior outage and notice that the
voltage was actually .6 volts, and you could have a negative
growth rate.
How we address that in the methodology, to be
conservative, is, we make the utilities assume that the
growth rate was zero. So all negative growth rates in the
probabilistic analysis for determining the end-of-cycle
distribution are assumed to be zero.
DR. CATTON: Because it's physically impossible?
MR. KARWOSKI: It's physically impossible.
DR. CATTON: But you don't do that with the
iodine?
[Laughter.]
DR. CATTON: Just thought I'd mention that.
Between two cycles, I can see how you're going to
predict the third. Do you ever go back and make a
comparison of what was anticipated and what's measured?
MR. KARWOSKI: You're getting to the punch line.
I will present that. That's a very important aspect of this
methodology, very important.
So, with the beginning of cycle voltage
distribution, and the growth rate distribution, I also need
NDE uncertainty. The NDE uncertainty stems primarily from
two distributions.
I only have a picture here, but there are two
distributions that get sampled. The first is analyst
variability, and that's a result of two different analysts
using the same procedures, could look at the indication and
get different voltages.
There was a study performed by the industry when
this methodology was being developed, which indicated that
basically the mean of the distribution was zero, it was the
normal distribution. I believe the standard deviation was
about ten percent.
The important point is that they recognized that
different analysts can call an indication different. There
is an analyst-variability portion to the NDE uncertainty.
The other portion of the uncertainty is
uncertainty associated with the wearing of a probe. If a
probe wears, it will be a different distance away from the
tube, and it could result in a different voltage.
As a result, the industry did some tests to assess
the effect of probe wear, and there's another distribution
which takes into account, the wearing of a probe.
So those are the two components of the uncertainty
distribution.
To arrive at the end-of-cycle --
DR. KRESS: Are those distributions clad randomly
to the original distribution?
MR. KARWOSKI: Yes. You basically --
DR. KRESS: Sort of like a Monte Carlo.
MR. KARWOSKI: It's a Monte Carlo simulation. You
sample the beginning-of-cycle voltage distribution, you
sample growth, you sample the two NDE uncertainty
distributions.
DR. KRESS: Do that over and over till you get a
new distribution?
MR. KARWOSKI: Right. And the number of
indications that you leave in service is a result of
whatever the -- you take your detected, divided it by .6,
you will grow everything that exists, and you will also have
the indications that you missed.
The one other factor, the .6 does not only account
for indications that were missed during inspection, it also
accounts for indications that could develop over the cycle.
Because we do not explicitly say there could be 50
more indications in the steam generator, the .6 POD accounts
for both those factors, the missed indications and
indications that can initiate during the course of the
cycle.
DR. KRESS: What do you do with this bottom
distribution, once you get it?
MR. KARWOSKI: With the end-of-cycle, that's my
next slide.
DR. KRESS: Okay.
DR. BONACA: I would like to know, how does it
account for those that are not detected? I mean, the .6, if
I understand it, you take the reading, which is the voltage,
and you multiply it by 1.6 or whatever, or you divide it by
.6 to get the new voltage value.
MR. KARWOSKI: To get the number of indications.
If you found one indication, and it was below the repair
limit, and you could leave it in service, you would take the
one indication, divide it by .6, and you'd end up with
roughly 1.7 indications.
You would take that 1.7 indications that would be
at that specific voltage, because you're not repairing
anything. You'd sample 1.7 indications and propagate it
through.
DR. BONACA: So the number of indications that you
are separating then?
MR. KARWOSKI: Right. It's the number of
indications at that specific voltage. So, recognize that
one of the industry complaints of our model is they say that
as the voltage gets -- as the voltage rises, they believe
their probability of detection increases, because it's a
much bigger flaw, the noise or the signal will come much
clearer out of the noise and the analyst isn't going to miss
it.
So, one of their criticisms is this constant POD
model. If I had a 13.7 volt indication, and I detected it
and I'm going to repair it, this model will require for them
to leave in at the beginning of cycle, 7/10ths of an
indication that is at 13.7 volts.
Their criticism is, if we had a 13.7 volt
indication, we would find it. Their position is 100 percent
of the time. We say you need to leave .7 of an indication
there.
DR. KRESS: When you divide by the .6 and get a
number, the number of indications is an integer, a whole
number. Do you round it up to the next whole number, of
just leave it as a fraction?
MR. KARWOSKI: The fractions are propagated.
DR. KRESS: The fractions are propagated, okay,
which is all right when you're doing a Monte Carlo.
MR. KARWOSKI: But, yes, you're right. But if
you're counting multiple bursts or something, it poses some
challenges, but we leave all the fractional indications in
service.
So what do we do with that end-of-cycle
distribution? I said there's two things we do:
We do the probability of rupture calculation, and
we do the conditional leak rate calculation.
With respect to the probability of rupture, we
start with this end-of-cycle distribution. We have our
burst pressure correlation that we discussed this morning.
It has scatter around it. We sample around it.
If we picked ten volts, we'd come over and say that for a
ten-volt indication, what is the range of burst pressures
that we can have?
We take that sample. Then we say this has been
normalized to a specific material property. We then come in
here and take a sample of our material properties
distribution, scale the burst pressure, determine what that
burst pressure for that one indication is, and determine
whether or not it's going to rupture under steam line break.
We repeat the process for all of the indications
in the steam generator and determine if there was a rupture
in that steam generator during that one Monte Carlo cycle.
Then we repeat it, tens, hundreds, thousands of
times to determine the probability of rupture under steam
line break conditions.
Leak rate calculation --
DR. CATTON: And this is done with 100-percent
evaluation of the steam generator, all the tubes are
checked?
MR. KARWOSKI: All the tubes are checked, 100
percent at each intersection.
Another conservatism in the model is there can be
multiple indications in a tube. The model treats them all
as if they were independent tubes, so if you had two
indications in the same tube, theoretically you could get
two bursts from that and it's just counted -- they're
counted as two multiples.
DR. CATTON: The multiples can be greater than the
number of tubes?
MR. KARWOSKI: Yes, yes. That's usually not the
case, but in the extreme, you're correct.
DR. SIEBER: Just so I understand, it seems to me
that if you repaired everything that you postulated would
leak, the fact that you end up with -- break, excuse me --
the fact that you end up with a probability of burst, really
comes from all these uncertainties that you have factored
into this.
Otherwise, you would know it perfectly that
everything would be accurate, and you could fix everything.
MR. KARWOSKI: Right, and that's one of the
industry's criticisms of our model, because even if they
repair everything, because of that probability of detection
adjustment and the uncertainties, they can predict extreme
--
DR. SIEBER: There's going to be something in
there?
MR. KARWOSKI: Right.
DR. SIEBER: Okay, thank you.
MR. HIGGINS: So you could you explain the
probability -- after you do the probability of rupture
calculation, is there an acceptance criteria there that
would then cause them to go back and make additional repairs
beyond what the voltage requires them to make?
MR. KARWOSKI: Yes. The acceptance criteria --
there's a reporting requirement. If the conditional
probability of burst exceeds one times ten to the minus
second, then they are required to notify us.
One of the corrective actions would be to plug
more tubes. Recognize, though, the other problem with --
not problem -- one of the issues with the methodology is
that you could start off with a very high probability of
rupture because of that POD adjustment.
MR. HIGGINS: And has that happened? What are the
typical results?
MR. KARWOSKI: Typically, usually, the few high
voltage indications dominate the burst probability, and so
once you start leaving 7/10ths of an indication in service,
that drives the probability.
MR. HIGGINS: I mean, have you had cases where you
had to repair additional -- plug additional tubes beyond
those that would have been called for by the one volt or two
volt criteria?
MR. KARWOSKI: I don't believe that has happened,
although I think that in one instance -- there is one
instance that I'm definitely aware of where the probability
exceeded one times ten to the minus two.
I believe it was 1.2 times ten to the minus
second, so I know there was one instance where it went over.
DR. SIEBER: The limit that you're looking for is
one times ten to the minus two?
MR. KARWOSKI: Yes.
DR. SIEBER: Okay, thank you.
DR. KRESS: Where did that number come from?
MR. KARWOSKI: That is on a slide towards the
back, but it's 1/5th the value that was assumed in the
staff's assessment of risk assessment of steam generator
tubes in NUREG 0844.
The value was just 1/5th because this is only one
degradation mechanism. We assumed a probability of rupture
in that report of five times ten to the minus second and we
said we didn't want one mechanism controlling, you know,
right up to the limit. There are other mechanisms going on.
The other reason is that it gives us some insights
on whether or not any tubes may not meet the Regulatory
Guide 1.121 structural criteria. Even though we're
calculating the probability of burst, we wanted some
insights on whether or not there are any tubes that are
starting encroaching on the deterministic structural
margins, the worst case tube.
DR. KRESS: Where did the number -- what's the
technical basis for the number as multiplied by five?
You say you took one-fifth of the number. What's
the technical basis for the --
MR. KARWOSKI: For the five times ten to the minus
second?
DR. KRESS: Yes.
MR. KARWOSKI: Somebody else -- basically I think
they did a risk assessment that assumed a frequency of
rupture of five times ten to the minus second, propagated
that through a risk assessment, and determined that that was
acceptable.
That's my understanding.
MR. STROSNIDER: That's correct. The assumption
in NUREG 0884 --
MR. KARWOSKI: 0844.
MR. STROSNIDER: 0844, I always get that mixed up.
But the assumption was made of a conditional failure
probability, given a main steam line break, of five times
ten to the minus second.
And when that was worked through the whole risk
assessment then, it was found that it gave an acceptable
level of risk.
All right, and then we reduced it to account for
the potential for other modes of degradation. I think Steve
--
DR. KRESS: It gave an --
MR. STROSNIDER: -- might be able to explain that
in more depth.
DR. KRESS: Okay.
MR. LONG: I wasn't involved in 0844, but Emmett
can correct me if I get this wrong. Initially, they were
trying to determine what they thought the conditional
probability of burst would be for a main steam line break,
based on experience, and largely that was experience of
ruptures that had occurred and some estimate of the period
of time that it would take to grow from where they could be
susceptible to rupture when the depressurization occurred,
to the period when they just ruptured during normal
operations.
So there was sort of an exposure estimate in
there. And I think it was essentially that exposure
estimate, with some adjustments for things they didn't think
they really observed, it was put into the risk assessment in
0844 to see if there was an acceptable or an unacceptable
situation.
It wasn't intended to be a limit when they did
that calculation. And it wasn't a risk assessment that
recognized severe accident sensitivities or that sort of
thing.
So it was done about 1985, I guess.
MR. MURPHY: The report was issued in 1988.
MR. LONG: So that's sort of the first of three
risk assessments that you will hear about, the one for 1477
being the next, and 1570 being the most recent one.
DR. KRESS: We'll hear about this later, will we?
MR. LONG: To some degree. We didn't intend to go
into it that way, so that's why I'm explaining it now.
The point I was trying to make, though, is, we
didn't try to use the .05 as some sort of an acceptance
criteria. It came from experience, and it was essentially
evaluated to see if it was something that we should try to
backfit.
MR. STROSNIDER: But I might add that when we were
developing Generic Letter 95-05, that's the risk assessment
that we had to look at.
And when we looked at it, we said, well, five
times ten to the minus second, when that was propagated
through the risk assessment for those sequences associated
with main steam line break, it resulted in an acceptable
level of risk.
So, we said, okay, we'll use that as an acceptance
criteria.
DR. KRESS: Yes. What was your criteria for an
acceptable level of risk?
MR. STROSNIDER: I don't know what it would have
been in '88 in terms of -- we'll have to get back to you,
okay?
MR. KARWOSKI: The one other item that I should
point out with respect to the probability of burst
calculation is that it's evaluated at a 95-percent
confidence value.
It's not just whatever the probability is; it's at
a 95-percent confidence value.
The other portion, the leakage integrity portion,
is depicted on this viewgraph. It's a similar methodology;
you do a Monte Carlo analysis. You sample the end-of-cycle
distribution for a specific voltage.
You come to the probability of leakage
correlation, and determine whether or not the tube will
either leak or it won't leak. If it leaks, you come into
the leak rate correlation and determine --
DR. KRESS: If you have a probability of leakage?
That's like Schrettinger's cat; it's both dead and alive.
What do you mean, it won't leak or it will leak? If it has
a probability of leakage, it has a probability.
I didn't understand your statement; that's what
I'm saying.
MR. LONG: Oh, for a given voltage, there is a
probability that a tube will either leak or it won't leak.
DR. KRESS: It's still Schrettinger's cat. I
don't quite understand. If it's a probability of a leak,
then one minus that is probability that it won't leak. I'm
still having trouble figuring out what you're saying.
MR. STROSNIDER: The point is that the probability
depends on voltage.
DR. KRESS: Of course it does. But it's a
probability --
MR. STROSNIDER: But I think the point is that it
either will leak or won't leak. If it leaks, the
probability is one, and those are the datapoints that are on
the top.
If it doesn't leak, the probability is zero.
Those are the datapoints that are on the bottom.
DR. KRESS: Yes, but that's a delta function.
MR. STROSNIDER: I'm sorry, I didn't hear you.
DR. KRESS: That's a delta function; that's not a
distribution.
MR. STROSNIDER: Well, and this came up yesterday.
Art Buslick raised a question about how you fit the
distribution to those data.
DR. KRESS: Once you have a distribution, though,
it's just a probability.
MR. KARWOSKI: Let me try it this way: You pulled
these tubes, okay, and you test to determine whether or not
they're going to leak under steam line break conditions.
You've got various voltages, okay?
DR. KRESS: So you plot that.
MR. KARWOSKI: When you test this tube, it either
leaks or it doesn't leak; there is no -- it either leaks or
it doesn't leak.
So you have the voltage associated with that
indication, and you either know if it doesn't leak, or if it
leaks.
DR. KRESS: Sure.
MR. KARWOSKI: Okay? There may be five tubes with
the same voltage. Three of them may leak; two of them don't
leak, okay, for a given voltage, okay?
So when you come in there, say, you just had
indications of that voltage, roughly 60 percent of the time,
3/5ths, 60 percent of the time you will assume that that
indication leaks; 40 percent of the time, you will assume
that it doesn't leak. Does that --
DR. KRESS: Yes, that answers my question. It's
just a probability.
MR. HIGGINS: Are you talking now about applying
the methodology or the development of these curves?
MR. KARWOSKI: Also applying it. It's the
development of the curves and applying it, because once you
have this function -- and I'll talk about this in a minute
-- once you have this function, you have a relationship and
you can say that there is a certain probability that an
indication with that voltage will leak, and one minus that
is the probability that that indication will not leak.
So you go through this method; you sample your
voltage, you sample your probability of leakage to determine
whether or not the tube either leaks or it does not leak,
you determine the associated leak rates with that voltage.
You add it up for the entire steam generator, and you get a
value for leakage.
You repeat this for all the indications in the
steam generator, and you've got the one value of the leakage
for that steam generator, you repeat it, tens, hundreds of
thousands of times, and you will have a distribution of leak
rates.
You order those, you take the 95th percentile, at
the 95-percent confidence, and that's what you say your
leakage is under steam line break conditions. So it's a
95/95 leak rate, okay?
There was some discussion yesterday on the
probability of leakage correlation, and this one is not in
your handout. What I have plotted here -- and it's probably
not worth your effort to figure out which curve is which --
but it's basically looking at six different functions of the
probability of leakage.
Okay, there is no theoretical basis for the log
logistic curve that we are using. We've documented that in
NUREG 1477. There's no basis, theoretical basis.
DR. KRESS: And there's not enough data to
best-fit any of it.
MR. KARWOSKI: They all have equal -- so, why did
we choose the log logistic over any of the others? Well,
before I get into that, I mean, if you look at this some of
these functions -- and, truthfully, I don't even know which
ones the log logistic predicts a probability of leakage for
very small indications, most of them don't.
These curves criss-cross. To determine which one
is conservative, you're not going to be able to do it,
because it's going to depend on the distribution of
indications you leave in service.
You can do it only if you analyze that steam
generator, because these curves are criss-crossing, and so
it depends if you have a lot of low voltages, medium
voltages, or high voltages.
In NUREG 1477, we did some analysis where we
shifted the distribution and determined the leak rates. And
in some cases, I believe it was the log co, she was
conservative, and in other cases, the log logistic was
conservative.
So, the staff did look at, you know, should we be
using a different function or how should we implement it?
What the staff subsequently decided was that the
log logistic was acceptable for several reasons, one being
the POD adjustment was conservative; we're evaluating leak
rate at the 95th percentile and a 95-percent confidence.
Those are two primary reasons, so the staff did
look into which curve to use, and we chose the log logistic.
The next few slides just basically describe what I said,
conditional probability of burst. It discusses the process
and the acceptance criteria. And this slide discusses the
leakage distribution.
I wanted to spend a little more time on here
because I think, as Dr. Powers pointed out, you know, it had
very low correlation coefficients for some of those leak
rate correlations, and there has been considerable study of
this data.
The way the methodology works for the leak rate
correlation -- and I don't believe this one is in your
package. This is something that I prepared subsequently.
DR. KRESS: No wonder I couldn't find that.
MR. KARWOSKI: If the linear correlation can be
developed between the leak rate and the voltage -- this
isn't the actual data, so you have the data in those
proprietary summaries that I passed out -- but -- well, let
me start over.
When you look at this data, the original data, you
could look at it and say, well, should the curve be like
this? Should it be like this? Should it be like that?
The industry did a linear regression and came up
with a curve. The staff was concerned, is there really a
correlation there?
We had a statistician look at it. And basically,
what he concluded, at least for the 7/8ths inch diameter
database -- and it may have been the same with the 3/4 inch
at the time, but for the 7/8 inch database, he concluded
that there wasn't sufficient confidence that the slope was
not zero.
So, basically he said there is no correlation.
The implications of that to the utilities is that a tenth of
a volt indication, the way we interpreted that is,
regardless of the voltage, the indication will leak by the
same amount.
And that's what the Generic Letter describes, an
acceptance criteria for showing whether or not you have a
correlation or you don't.
And it's a standard statistical test, a P-value
test, and you have to have a 95-percent confidence that the
slope of the line is not zero.
DR. KRESS: This is sort of a generic correlation
that you build up a database for.
MR. KARWOSKI: Right.
DR. KRESS: And you're saying that before -- I'm
not quite I understood. You're saying that before you can
use this in your alternative criteria, you have to have
enough data to have a correlation?
MR. KARWOSKI: No.
DR. KRESS: I'm not quite sure what you're saying.
MR. KARWOSKI: What I'm saying is, if you can just
demonstrate by statistical tests that the slope of this line
is non-zero --
DR. KRESS: I thought you already said that you
tried that, and it wasn't.
MR. KARWOSKI: The Generic Letter is more
performance-based. Instead of saying here is the
correlation at this time, and there is the slope, we
recognize that they're going to pull additional data.
They're going to get more data which --
DR. KRESS: At some point in time, you may have a
correlation?
MR. KARWOSKI: That's correct, or you may go from
correlation to not. The bottom line is, when you get more
data, you have to put it in these correlations and you have
to use the most recent database.
We didn't want to lock in a certain database,
knowing that utilities will gain more experience.
DR. KRESS: Up to the point where they
statistically can't say they have a correlation, what do
they do up to that point?
MR. KARWOSKI: When they don't have a correlation,
they have to assume that essentially the tube will leak at a
specified value.
DR. KRESS: And that value is what?
MR. KARWOSKI: They basically do a Monte Carlo
analysis and model the error and say -- around the
regression line, what the slope at zero. So basically they
average the log of the leak rate, obtain a value, and then
sample the uncertainty around that line, and basically
assume that the leak rate is independent of voltage.
A tenth of a volt indication will leak the same as
a 25 volt indication.
DR. KRESS: Put that line through the average of
the data?
DR. CATTON: That's what it amounts to.
MR. KARWOSKI: Through the average of the log, it
will leak -- they basically do a -- right.
DR. KRESS: That doesn't sound like a very
interesting thing to do.
MR. KARWOSKI: Well --
DR. CATTON: It's zero physics. It's simple, too.
DR. BALLINGER: But what is the criteria, exactly,
by -- what is the fence that they get over before they say
it's a correlation? What is the statistical test that has
to be passed?
MR. KARWOSKI: The P-value has to be less than
.05. There has to be a 95 -- I hope I get this right; I'm
not a statistician.
It has to be a 95-percent confidence that the
slope of that regression line is not zero.
DR. KRESS: That's the sort of standard rule of
thumb for statisticians, I guess.
DR. CATTON: If you don't know a damn thing, you
just average everything.
MR. KARWOSKI: Well, not only do you average it,
but you also model the uncertainties.
DR. CATTON: The only thing you're bringing in
over here is the distribution of voltages. A tenth of a
volt is probably not going to show -- you just toss out all
the physics when you do that.
MR. KARWOSKI: This is an empirical --
DR. CATTON: It's not even empirical. I mean,
you've just thrown everything away.
DR. KRESS: There's no physics in there anyway.
DR. BALLINGER: But it just says there's no
information in the data.
DR. CATTON: That's right. So how do you pick the
number that you use that's meaningful?
DR. KRESS: I think that's your point, right?
DR. CATTON: That's right, how do you know?
MR. STROSNIDER: This is Jack Strosnider. I'd
suggest that one of the things that might need to be pointed
out is that there is a very large uncertainty -- not
uncertainty but actually variability in scatter and leakage
rates through cracks.
And we're talking orders of magnitude, as I
recall, three leakage values. And that's part of what
drives this, is there is a large variability.
And there was discussion yesterday about what sort
of particulate might be in the cracks, and what the
morphology is and that sort of thing.
DR. CATTON: That's right.
MR. STROSNIDER: And even when you grow single
stress corrosion cracks in a tube, you get very large
variability. So trying to get a lot of information out when
you've got that much scatter, that may contribute to some of
the problem here.
But anyway, that's -- the industry was not happy
with what was sometimes referred to as the flat earth model,
where we said, okay, just take this horizontal line.
But as Ken indicated, we tried to build into it,
as the database grew, if the correlations became more
obvious, then it could be used.
DR. KRESS: We have an imminent --
MR. SHACK: This is Bill Shack. Let me just take
a crack at it. This is a bad joke here.
DR. KRESS: Fine.
MR. SHACK: One of the reasons that -- you know,
physically, I think this thing sort of works out the way you
expect it to work out.
There is a rough correlation for cracks up to a
half an inch between the crack length and the voltage, that
is, the voltage increases rather rapidly with crack length
up to about half an inch, and then it kind of flattens off.
Well, most of the cracks of interest here are on
the order of something less than a half an inch, so that an
increase in voltage is a measure of crack length.
Burst pressure depends only on crack length. So,
I would expect to get a reasonable correlation between burst
pressure and voltage.
Leak rate doesn't depend just on crack length; it
depends on crack depth. And so I expect to get a hell of a
lot more scatter in my voltage versus leak rate correlation,
than I do in my voltage versus burst correlation, and so I
get a better statistical fit for the pressure, burst
pressures, than I do for the leak rate, because I'm trying
to look at both.
And as Jack said, even if I knew the depth of the
crack and the length of the crack, I still have more
uncertainties in the leak rate, because it depends on lots
of things.
DR. KRESS: Don't you have to have a through-wall
crack to get a leak rate?
MR. STROSNIDER: Right, and that's probably the
biggest thing here, but that's sort of accounted for in the
zero to one.
Either I've got a through-wall crack or I don't.
DR. KRESS: That's sort of wrapped up in that
distribution?
MR. STROSNIDER: That's wrapped up in that
distribution. But then once I do that, even though I have
decided I've got a through-wall crack, I really don't know
how much the through-wall length is, where the burst
pressure is.
I just keep upping the pressure until I rip the
thing open, and I know really how long it was.
All I'm arguing is that the fact that you have
lots of scatter here and you get a better correlation, the
burst pressure, isn't a surprise. But the answer is, you
shouldn't expect a terribly good correlation here, so what
are you supposed to do?
DR. CATTON: And it's a nice physical explanation
of the observations. But the question still remains, what
are you supposed to do?
I guess the only thing is that you have a low-end
cutoff on the voltage; don't you, with that probability of a
leakage. If the voltage is less than some number, it
doesn't leak.
So when you say even a tenth of a volt has a leak,
it doesn't, because there is a cutoff.
MR. KARWOSKI: Right, wherever there is some --
DR. CATTON: There's a cutoff. And I bet that
you're at zero at a tenth of a volt.
MR. KARWOSKI: You could be. I don't recall.
DR. CATTON: That's what that is, it's a gate that
sits right in the middle.
MR. STROSNIDER: This is Jack Strosnider. Art
Buslick could address this better than me, but I think all
those fits go through zero. But the probabilities do become
very, very small.
DR. KRESS: For the log, the logistic, you don't
have a cutoff; it goes to zero asyntotically. There is
still a small probability.
MR. STROSNIDER: It could be a small number, but
the utility, the thing they didn't like about it is that if
they had a large number of small voltages, that they could
still generate some sort of leakage.
DR. CATTON: I would be concerned if I were them,
too.
But from the other side, you'd probably get a
too-low rate at the high voltage.
DR. KRESS: Basically what they're saying is that
those two sets of data on the top and bottom, and when you
get down that low, there are no overlapping, although this
curve would have said there was some overlapping.
MR. STROSNIDER: I think the answer to your
question with what you do with it -- and this may just spur
more discussion -- but my response is that you take a
conservative approach, which is what we think we did, in
terms taking the constant leakage --
DR. CATTON: Where you put the line is on the high
side, rather than the low side?
MR. STROSNIDER: Well, Ken can explain it.
DR. KRESS: I think we may have to.
MR. STROSNIDER: When you go through this
evaluation, then you look at some 95th percentile.
MR. KARWOSKI: If you look at the type of voltages
that the plants are running at, giving it the zero slope or
the flat earth, however you want to look at it, that's going
to be more conservative, because usually there aren't a lot
of large voltage indications.
What will drive the probability is the leakage
from those low voltages. You still have a finite
probability that those will leak.
MR. MUSCARO: Joe Muscaro with the NRC staff, I
just have a point of clarification. I think Bill mentioned
the phenomenon in reverse. I think he said that the leakage
depends on the length and depth and burst on length.
Well, it's the reverse, of course. Leak rate
depends on the length of the flaw, but burst depends both on
the length and depth.
The other point was the physical basis for whether
there should be a correlation or not with respect to voltage
versus crack length, which is important for leakage.
And there some studies we have done in the past
and have show that the voltage starts at a crack length of
about 3/10 of a inch, which makes this fairly important for
the application when we're talking about cracks that are
less than 3/4 of an inch long.
So there is a saturation effect on the voltage.
The physics limits it. I mean, it only sees a certain
portion of the crack, so there is no additional voltage
beyond about a 3/10 of a inch crack with the coil that we
use today. Of course, that's a function of the design of
the coil.
MR. HIGGINS: Ken, you said that this is being
updated, and how is that done? Is somebody the keeper of
the curve for all of industry, and then people use that?
MR. KARWOSKI: Essentially, that's correct, either
EPRI or NEI -- and I don't recall which one. I believe that
NEI sends in the formal report, but it's an EPRI report.
But basically, annually they update. There is a
specific protocol, and I believe it's annually. They
incorporate all the pulled tube data from that year, they
put it in the correlations, they develop revised
correlations, and they give those to the industry.
The case where I said that one plant exceeded the
one times ten to the minus two burst probability, what
happened there is, they did their inspection, they
calculated the probability of burst to be less than ten to
the minus two.
They incorporated more data into the burst
pressure correlation, and when they did that and reran their
burst results, they ended up slightly greater than one times
ten to the minus second. That's what happened in that case.
So, utilities, when they do these calculations are
supposed to use the most recent database.
DR. CATTON: Is this number increasing or
decreasing with time?
MR. KARWOSKI: The?
DR. CATTON: The leak rate, the uncorrelated leak
rate number?
MR. KARWOSKI: I believe there is a correlation
now for both the 3/4 inch and 7/8 inch database.
DR. CATTON: So it's beginning to get a little bit
of a --
MR. KARWOSKI: A slope, right.
DR. KRESS: So at some magic point in time, people
will quit using this flat earth average with the
distribution around it, and go to this actual curve?
MR. KARWOSKI: With also distribution around it,
because as was pointed out, there is a lot of scatter.
DR. KRESS: And you won't be penalizing the small
cracks as bad?
MR. KARWOSKI: That's right, and it makes a --
once again, it's going to depend on your distribution of
indications, but for the one plant that I'm aware of, it
reduced their leakage by a factor of three.
They were predicting something on the order of 13
gpm, and when they went to this it was more like four or
five. So, it reduces the leakage considerably.
Okay, one of the important aspects of this Generic
Letter is that it requires the licensees to submit certain
data.
With that data, you can go back and compare the
projected and actual end-of-cycle distributions. That
allows you to check various things. Is the .6 POD, is that
accounting for new indications and the probability of
detection accordingly? Was your growth rate assumption
realistic?
Was my NDE uncertainty realistic? Because you can
compare what you actually found to what you projected.
That's a very important aspect of this.
These comparisons have generally shown the
methodology to be conservative. That's not to say that
they're perfect; that's not to say that they're haven't been
indications with larger voltages than we predicted, as was
pointed out yesterday. I will discuss that specific
example.
But the staff did do a comprehensive review of
what I have termed the 90-day reports, where the utilities
supply their inspection results, both their projections and
their actual results.
DR. KRESS: This is one of those cases where you
hope you're not gently flowing down the stream and suddenly
about to go over a waterfall, a cliff-edge type thing? I
don't know how you deal with that in terms of projecting
forward in time, based on past history.
MR. STROSNIDER: This is Jack Strosnider. I'd
just make a short comment on that, which is that I think
what you're talking about is how can I foresee what's going
to happen in the next cycle of operations?
DR. KRESS: Yes.
MR. STROSNIDER: That's an issue that exists,
regardless of what --
DR. KRESS: No matter what.
MR. STROSNIDER: -- what repair criteria or
inspection method you use. And you can go to laboratory
data to get some insights, all right, but we have found,
historically, we think, that the best predictor is what
happened in the last cycle.
That doesn't mean that you won't some day be
surprised with a new form of degradation or some higher
growth rate, and you have to deal with that when it occurs.
But nobody has a crystal ball to say exactly
what's going to happen in the next cycle.
CHAIRMAN POWERS: You're going to show us some
examples of where the methodology has been compared?
MR. KARWOSKI: Yes, I will show you -- I have two
examples. I just pulled out two.
But back in 1997, we did do a comprehensive review
of, I believe it was eight 90-day reports, and there were
some issues that were identified as a result of that.
One of the issues was, when you have a low number
of indications, your projections typically are off, but with
respect to the distribution that you find in the next cycle
-- but that necessarily is not a bad thing.
When you look it from the probability of rupture
or leakage, what you're finding out is, you know, you
predicted a ten to minus four probability of rupture, and
instead, it's five times to minus four.
In general, in all of these assessments, in not
one instance did we exceed the acceptance criteria of one
times ten to minus second for burst, or the applicable
leakage limit at that plant.
The other thing I'd point out is that as a result
of implementing this repair criteria, there has been no
significant operational leakage.
DR. CATTON: Your second paragraph says
comparisons have generally shown methodology to be
conservative.
MR. KARWOSKI: Yes.
DR. CATTON: What's your measure?
MR. KARWOSKI: The measure is several things: We
look at what the distribution of voltage is. Did we predict
the peak voltage?
Did we -- how did we do with respect to burst and
leakage? What was our predicted probability of burst and
what was our actual probability of burst, based on the
inspection findings.
DR. CATTON: I can buy into that, but now what
about leakage?
MR. KARWOSKI: The same thing for leakage: What
did we predict for leakage, and what will we predict, based
on our actual end-of-cycle findings?
DR. CATTON: End-of-cycle findings is a measured
leak rate for the steam generator? What is the end-of-cycle
finding that you compare with leakage?
MR. KARWOSKI: Okay, when you go in and do an
inspection today, okay, you're going to have a distribution
of indications. Before I do any repairs, I say, how much
leakage would I have gotten from that distribution of
indications?
DR. CATTON: Okay.
MR. KARWOSKI: Okay? Now, I go back to the prior
cycle. What did I predict I was going to get?
If I predicted --
DR. CATTON: But you haven't measured leakage and
compared it with prediction?
MR. KARWOSKI: No.
DR. CATTON: So this is all just paper? It's an
earlier estimate with the present estimate, but there is no
actual comparison.
So you don't know whether you're conservative or
not. You may have been non-conservative in both cases, one
must more than the other or less than the other.
MR. KARWOSKI: The burst and leakage are a
comparison based on the methodology. The actual findings is
a comparison of what you have.
MR. STROSNIDER: This is Jack Strosnider. What
you have --
MR. CATTON: I thought I asked and you said no.
MR. STROSNIDER: What you have compared is you
have compared the predicted versus the measured end of cycle
voltage distribution.
MR. CATTON: Yes.
MR. STROSNIDER: That is what you have compared.
In terms of what leakage would be associated with
those two distributions, it uses the same methodology. As
you say, if there is some bias in that methodology it would
be there in both --
MR. CATTON: So you really don't know where you're
at.
DR. KRESS: There's no additional information --
MR. CATTON: That's right.
DR. KRESS: -- in going to that second comparison,
yes.
MR. CATTON: So this second statement, comparisons
have generally shown methodology to be conservative we don't
know.
MR. KARWOSKI: In predicting the end of cycle
voltage.
DR. KRESS: With respect to that you can say that.
MR. CATTON: It only has to do with voltage. It
has nothing to do with the leakage.
MR. KARWOSKI: Yes.
MR. CATTON: We have a model that we might
believe --
DR. KRESS: You're comparing the bursts --
comparing the voltage and then comparing the leakage tells
you nothing, no additional information.
MR. CATTON: Bursts I buy because they actually
have correlations and so forth but the leakage is so all
over the map I don't think you can come to a conclusion.
DR. KRESS: But even comparing -- you know, it
makes no sense to compare predicted bursts to actual bursts
or predicted leakage to actual leakage. The comparison you
are getting is voltage to voltage, predicted versus actual,
and that is the only information you have really.
MR. CATTON: That's right.
MR. HIGGINS: But the leakage database should be
getting better as you go over the years because you are
adding more and more pulled tubes that you are testing.
DR. KRESS: Yes. Yes, the database ought to get
better. Your ability to convert the number in the leakage
is getting better.
MR. STROSNIDER: I think I'd just comment I think
it's kind of indicated there is a correlation in the leakage
database now -- you know, because additional data have been
added and it meets the criteria that were established but
again I would come back and make the point again that when
we start talking about leakage under main steam line break
conditions that we are always relying on a model.
DR. KRESS: Oh, yes.
MR. STROSNIDER: Okay? We have a model here that
is tied to voltages. If you had crack measurements that
were accurate that you believed you would be going to
something like CRACKFLOOR or one of the other models that
was discussed yesterday, so yes, we are constrained by
having a model. That is just a reality of it.
DR. KRESS: These leakage correlations were
developed using the normal operating delta p, right?
MR. KARWOSKI: No, these are steam line break --
DR. KRESS: Those are steam line break leakages
that used the actual steam line break delta p.
MR. KARWOSKI: Right, and as I mentioned this
morning there are some adjustments to that data.
The test procedure calls for them to be run at
operating temperature. There are some data -- I looked at
lunch -- there are some data that are at room temperature
and they are adjusted over here.
The delta p could differ. The delta p for the
test --
DR. KRESS: You might not have tested at the delta
p --
MR. KARWOSKI: At 2650 -- it could be 2500, 2400.
I would have to look at each individual datapoint but there
is a range of differential pressure.
DR. KRESS: They would have corrected for that
some way?
MR. KARWOSKI: They would have corrected --
DR. KRESS: If they knew how leakage varied with
delta p they would have corrected maybe.
CHAIRMAN POWERS: They would know leakage varies
with delta p?
DR. KRESS: I don't, do you?
CHAIRMAN POWERS: I thought you knew everything.
DR. KRESS: If they were flowing through a pipe
I'd know.
[Laughter.]
CHAIRMAN POWERS: So how can they do the
correction?
MR. KARWOSKI: I am not aware of all the details
with respect to that correction procedure. I know that they
have corrected. I don't know the specific details of that.
We would have to get back to you on that.
The procedure is documented in the EPRI test
procedures.
MR. HOPENFELD: I'd bring to your attention that
this was a major point that I was making yesterday and I
wouldn't let you out of this room until you answered it.
MR. STROSNIDER: I think that the Staff committed
to get back on that subject and we will. Thank you very
much.
MR. HOPENFELD: Okay.
CHAIRMAN POWERS: I regret that I was called away.
Have we had a chance to discuss the famous log logistics
question.
DR. KRESS: We passed right through it just like
that. You want to go back to it?
[Laughter.]
MR. BALLINGER: Not without getting a definition
of what they defined as being proof.
CHAIRMAN POWERS: Okay, but sometime I would like
to hear about the log logistic curve just because I happen
not to know what a log logistic curve is.
Please continue.
DR. KRESS: We decided it was conservative for
very small cracks because it doesn't have a bottom cutoff
for it.
MR. KARWOSKI: Here is one example of a comparison
of actual and predicted. I just -- I don't want to spend a
lot of time on this one. I would rather go to the following
one that was discussed yesterday.
Basically if you look here the actual is the open
and the predicted with the POD of .6 is in black. This
other one is an approach proposed by EPRI that the Staff
hasn't approved, but if you look at this, here is the actual
number of indications predicted in this voltage range versus
the predicted and in general although you can't see out
here, in general we have been conservative.
CHAIRMAN POWERS: Maybe I need a little more
explanation.
You have plotted a number of indications versus
bobbin voltage and the black bars are what you predict and
the open bars are what was actually measured --
MR. KARWOSKI: Right.
CHAIRMAN POWERS: -- and when I look at the lower
areas you get more low voltage indications than you predict
but as you move up the voltage then you predict more than
you actually observe. Is that correct?
MR. KARWOSKI: But there are exceptions but in
general you tend to be conservative in this. This was just
one example. You are right. There are exceptions and the
Farley example
MR. STROSNIDER: Ken, I just might point out that
I think it gets back to using the leakage model, okay?
When you ask yourself is that distribution
conservative or not, what you do is use the predicted versus
the actual, calculate the leakage and say does this end of
cycle distribution result in more or less leakage than would
have been predicted.
Again, I think Ken said earlier that when we do
that in every case the leakage has been bounded, is that --
MR. KARWOSKI: What I said is that in every case
when you do that the leakage is always less than the
acceptance limit based on the dose equivalent of Iodine 131
in the tech specs.
In some cases it may have been higher than the
predicted but it has always been less than the dose allowed,
the leakage that will be permitted --
CHAIRMAN POWERS: Do you have the numerics on
that?
MR. KARWOSKI: Numerics? Like the number of times
it's happened or --
CHAIRMAN POWERS: Well, like a leakage predicted
versus a leakage actual -- I'm sorry. A leakage predicted
versus leakage predicted from the actual distribution?
MR. KARWOSKI: We have that data. I think I may
have it for the Farley.
CHAIRMAN POWERS: Oh, good.
MR. KARWOSKI: But I don't know if I wrote it
down.
This is the Farley example. Getting at the
comparisons, the statement that the comparisons have
generally been conservative, this is an excerpt from a
Farley submittal. This would be their end of Cycle 14
projections, what they planned on, what they expected to
find.
Here's the number of indications that they
expected. Here's the maximum voltage that they projected.
Here's the burst probability, single tube burst probability
and here is the steam line break leak rate, okay?
If you look, let's look at the number of
indications. In all cases the projections bounded the
actuals. The max voltage, in this case they are equal.
The projection, in Steam Generator B the
projection was greater. In Steam Generator C this is the
indication we were talking about yesterday. It was 13.7
volts but this value is adjusted for NDE uncertainty and
that is why it is listed as 14.9 but it is the same
indication.
In this case we underpredicted the max in the
cycle voltage.
With respect to burst probability, if you look at
these values --
CHAIRMAN POWERS: These means or are these the 95
percentiles?
MR. KARWOSKI: These are the 95 percentiles.
Steam Generators A and B we conservatively
predicted and the burst probability Steam Generator C we
underpredicted.
What I was saying was this value was still less
than 10 to the minus 2.
I would also point out that what is driving this
burst probability is that single, is that large voltage
indication.
With respect to leakage with Steam Generators A
and B our projections were conservative compared to here.
Likewise for Steam Generator C.
CHAIRMAN POWERS: Why was that?
MR. KARWOSKI: It was probably because they were
using the leakage correlation which had a zero slope. That
is probably what was causing that. I don't know that, but
that would be my guess.
MR. BALLINGER: Do you have information on what
the end of Cycle 13 actuals were?
MR. KARWOSKI: We would have to dig it up. I
don't have it here.
DR. KRESS: Is there a consistent underprediction?
MR. KARWOSKI: I tried --
DR. KRESS: For two cycles or --
MR. KARWOSKI: During lunch I tried to get that
information because the end of Cycle 15 has already
occurred.
From what I was told and I have not reviewed this,
from what I was told this did not occur in the next cycle.
I would need to review that and provide that. I tried to do
it during lunchtime. I just ran out of time.
MR. STROSNIDER: Again I think it might be worth
pointing out too that in this methodology that 14, almost 15
volt indication would get folded back into the prediction of
the next cycle's end of cycle distribution, all right? So
the fact that that occurred does influence the assessment of
the next operating cycle.
MR. KARWOSKI: Absolutely and if you look at the
end of Cycle 15 projections for Steam Generator C, your
burst probability of 9 times 6 times 10 to the minus 3 is
dictated by that seven-tenths of an indication that you left
in there at roughly 13.7 volts.
DR. KRESS: Will they repair that one or --
MR. KARWOSKI: Absolutely. That indication was
actually --
DR. KRESS: So if they repair that one, how do you
fold it back into your projection for the next --
MR. KARWOSKI: Oh, because that's -- you just do
that on paper. You plug a lot of tubes but you still
include them in your analysis -- through .6 POD so --
DR. KRESS: Okay. They can never escape that .6.
MR. STROSNIDER: You are assuming that you missed
another one of those at some probability.
DR. KRESS: You never escape that -- that's right.
CHAIRMAN POWERS: You have an acceptance value for
the burst probability. What about the leak rate?
MR. KARWOSKI: The leak rate is based on the dose
equivalent Iodine 131 and whatever value that have in the
tech specs there's associated leakage which they --
CHAIRMAN POWERS: Yes, I was asking do you know
what the number is --
MR. KARWOSKI: For Farley?
CHAIRMAN POWERS: We are asking you a lot.
MR. KARWOSKI: I looked at all this and I have it
written down in my office. I don't have it here but I think
it was -- I think the value was on the order of 11 GPM,
something like that. I would have to look it up. There's a
history associated with that.
DR. KRESS: Let me ask you about this --
MR. KARWOSKI: I can tell you this value is less
than whatever was in the tech spec at that time.
CHAIRMAN POWERS: That is really what I wanted to
know.
DR. KRESS: Let me ask you about these phantom
particles -- I mean phantom cracks -- regarding the .6.
Let's say I have this 14.9 voltage indication and
I say all right, I am going to repair that one but I have
got another one in there at .6 that I am going to carry to
the next cycle and predict the voltage and the leakage.
If I go to the next cycle, if I go in and make a
bunch of measurements, now I don't see this 14.9 again
because I prepared it and I really didn't have that phantom
crack that I thought I did. Do you now throw it away, the
phantom crack? Do you throw that away although --
MR. KARWOSKI: The growth rate in this may include
remnants because when you do the growth distributions you
have to take the most conservative of the two consecutive
cycles, so there may be a part of that 14 volt indication
hidden in the growth, because in this case you had like a 12
volt --
DR. KRESS: So you don't carry these phantom
cracks forever?
MR. KARWOSKI: No, they eventually will disappear.
MR. HOPENFELD: May I ask Dr. Powers a question?
I think in a case of Farley I think originally
they didn't have iodine spike in their tech specs and they
didn't take it into account, and I think it was something
like 60 and then it was pointed out to them that they had to
take iodine spike into account they backed up to whatever it
was, something like 20 or something -- I think 20 to 30,
roughly.
MR. KARWOSKI: That specific tube -- this
viewgraph which isn't in your handout also basically just
summarizes the Farley and the Cycle 14 results. Tube was
pulled. Predictions bounded the actuals except for Steam
Generator C. We discussed the two exceptions.
This basically says that they were still within
their limits for burst and leakage.
The tube was pulled for destructive examination.
It grew from 1.4 volts to 13.7 volts so 12.3 volts changed
during the cycle. All the voltages used in the correlation
are prepulled voltages. When you pull these tubes if there
is any cellular component it can open it up and change the
voltage. It can cause some damage to the crack.
The post pull voltage was 28 and a half volts. I
just point that out just to show that there may have been
some damage.
This tube burst, slightly higher than 1.4 steam
line break, which I believe would be expected based on the
correlations. This tube also leaked at three-quarters of a
gallon per minute, and the other observation I would make --
they put an in situ pressure test and tried to leak test
this while the tube was still in the steam generator so the
support plate is still there. It's still packed with
corrosion products.
During that test there was no leakage but it did
leak in the laboratory test at three-quarters --
CHAIRMAN POWERS: If we had done that leak test
with a guy whanging on the support plate with a 16 pound
sledgehammer, would it have still not leaked?
MR. KARWOSKI: I can't say. All I can tell you is
the way we assume it in the methodology is that it leaks at
three-quarters a gallon per minute.
CHAIRMAN POWERS: But depending on crud and
what-not to prevent it from leaking may not be applicable
under a main steam line break.
MR. KARWOSKI: That's right, and our methodology
assumes, and I might not have pointed this out, all these
leak tests and burst tests take no credit for the support
plate. It assumes the degradation is entirely free-span.
CHAIRMAN POWERS: Okay, so this last line is just
a point of interest?
DR. KRESS: As a point of interest. It's a point
of a lot of interest.
When you take those tubes out, do you do anything
to them like clean them before you leak test them or just
take them over and leak test them?
MR. KARWOSKI: There is no cleaning --
DR. KRESS: No acid wash or anything like that?
MR. KARWOSKI: They take them from the field.
Whatever damage is done, you know, from the scraping through
the other plates and through the tube sheet they take it and
leak test it.
DR. SIEBER: Do they decon it at all?
DR. KRESS: That was really my question, decon.
MR. KARWOSKI: I don't know, to be honest.
DR. KRESS: Because that could be --
Do you have any idea why this thing went from 1.4
volts to 13.9 volts? That's that cliff edge I was worried
about.
MR. KARWOSKI: The tube was pulled, and they did
destructive examination. The tubes at the tube support
plate intersections at most plants have intergranular stress
corrosion cracking.
DR. KRESS: They did.
MR. KARWOSKI: When they did destructive
examinations of this tube, they noticed two things: They
noticed some transgranular stress corrosion cracking, and
they also looked at -- what they have said is that they have
noticed some fatigue type degradation at some fatigue
striations.
What they postulated is two things: The
transgranular stress corrosion cracking may have been
contributed to by the presence of lead in the steam
generator, which is a known mechanism.
Two, they think that they had done some pressure
pulse cleaning during the prior steam generator inspection,
and that this tube was at the location of where that nozzle
was, and so they believe that they may have propagated that.
That is what they have put in their reports.
DR. BALLINGER: This whole thing is predicated on
the fact that you don't introduce an additional degradation
mechanism into the system.
And what you're saying is that this datapoint is
exactly that. It's a datapoint for which you don't have
intergranular stress corrosion cracking and for which you --
that's not the mechanisms we're dealing with here.
This datapoint is basically useless.
MR. KARWOSKI: There was some transgranular
components. I didn't mean to imply that it was
predominantly intergranular stress corrosion cracking with
some -- I believe they said minor transgranular.
I did not look at it in any more detail than that.
The Generic Letter 95-05 approach has some
advantages: Basically, the licensees have to go in, do
their inspections, make projections to the end of the next
cycle, and then with those projections, you can go back and
say, well, how well was my methodology performing by doing
the -- when you do the next inspection.
And that's basically what condition monitoring is.
It's basically monitoring and evaluating the as-found
conditions in the steam generator, and comparing them to
what you predicted that you would have during the -- from
what you would have predicted from the prior cycle.
In addition, you take your as-found and you
determine whether or not you would have satisfied your burst
and leakage criteria during that cycle. And that's a
backwards look to ensure that you had adequate tube
integrity during the previous cycle.
CHAIRMAN POWERS: My understanding is that this
condition monitoring and operational assessment is a program
the licensees have committed to. It's not required by the
regulations, but they have just committed to it.
MR. KARWOSKI: They have committed to it, yes, and
to the NEI 9706.
MR. STROSNIDER: But in terms of -- and maybe
we're moving into the next assessment here, but -- or
subject.
In terms of Generic Letter 95-05, it becomes part
of their tech specs.
MR. KARWOSKI: Right. The utilities, as part of
95-05, have to submit certain information. That information
-- with that information, the staff can take a look at how
well the methodology predicts. It can determine what your
-- the staff can do that condition monitoring analysis.
The operational assessment is similar to all the
projections. Basically, you're taking a forward look and
saying will I be able to maintain tube integrity during the
course of the next inspection?
And it required knowledge of NDE uncertainties,
growth rate uncertainties, and the burst and leakage
correlations.
Before I talk about crack growth, I promised
earlier that I'd talk a little bit about the three-volt
alternate repair criteria.
Once again, this slide is not in your package, but
based on yesterday's discussion, I thought I'd clarify.
The three-volt alternate repair criteria was
implemented at one plant -- or at Byron I, one utility. And
it's a modification to the standard Generic Letter 95-05
approach. It's still a voltage-based approach, but it's
different.
What they did in this submittal is, they said we
want to take credit for our support plates being there. We
want to say that that degradation doesn't get exposed.
As a result, our probability of burst at that
location for axial indications will be minimal.
So what they did is, they went in and expanded
selected tubes above and below the tube support plate, and
essentially anchored the plate in place.
There were some small deflections. I think they
were intending to limit the deflection to a tenth of an
inch.
As a result, the probability of burst from an
axial indication is minimal, however, these larger voltages,
where you start to get concern then is axial tensile tearing
of these indications.
They developed another voltage correlation which
indicated that you needed voltages on the order of, you
know, ten to 20 volts. I don't recall the numbers, but it
would allow them to leave larger voltage indications in
service.
And the voltage limit basically is based on this
axial tensile tearing correlation, and that's where the
three-volt limit came from.
This approach isn't identical to 95-05 in terms of
the leakage assessment, because it introduces another
concern. If the tube can't burst, it may try to burst, but
once it hits the edges of the support plate, it will stop,
what they called indications restricted from burst or ERBS.
And they had to develop another leakage
correlation or another leakage database to say how much
leakage can I get from these tubes that attempt to burst but
don't fully open up?
And so they had to modify their leakage analysis.
I didn't want to go into all the details of this
methodology, but basically that's the three-volt criteria.
It's a different concern than the one and two volt
amendments of Generic Letter 95-05.
Byron and Braywood have subsequently replaced
their steam generators. No operating plants currently have
it, although I believe there is an application inhouse for
us to review this again.
CHAIRMAN POWERS: How long did Byron and Braywood,
how many cycles did they operate with that criterion?
MR. KARWOSKI: This amendment was approved, I
believe, in the '96 timeframe, so I'm going to say two
cycles. I'd have to look at it, but it's something like two
cycles.
CHAIRMAN POWERS: Is the application that you have
by another licensee to do that, is he also indicating that
he will replace his steam generator in short order?
MR. KARWOSKI: I don't know in that case. I know
that there is a different methodology. They're making
different assumptions, and the staff will perform a detailed
review.
DR. KRESS: There couldn't be a lot of database
behind those correlations.
MR. KARWOSKI: With respect to the leakage
database, they did an extensive test program as a result of
proposing this criterion.
DR. KRESS: They did?
MR. KARWOSKI: So they did many tests with respect
to how much leakage could I get, and they basically -- I'm
not sure they took the bounding leak rate, but they took
something near the bounding leak rate, and any tube that
they predicted will burst, you know, but contact the plate,
they would assign that limiting leakage to it.
MR. STROSNIDER: This is Jack Strosnider. Ken, I
think you're going to be going on to more general
discussions now with regard to crack growth rate?
MR. KARWOSKI: Yes, I was going to touch on that.
MR. STROSNIDER: I just wanted to make that clear,
again, that I think we completed the presentation on Generic
Letter 95-05 and the voltage-based criteria. And I just
wanted to make a clear demarcation here before we move on to
other issues, so they don't get confused.
MR. KARWOSKI: Okay.
MR. HIGGINS: Could I ask one final question on
that to maybe wrap it up in my mind?
Do you consider that these -- by applying this
Generic Letter 95-05, and applying these alternate repair
criteria, and then doing these special main steam line break
analyses where you assume that the tubes will be rupturing,
perhaps, or leaking, did you consider that for these plants
to be an accident of a new type and a change to the design
basis of the licensing basis for those plants?
And, further, do you consider that you've
addressed all of that by doing all these additional analyses
and tech spec modifications, et cetera?
MR. KARWOSKI: I'm not an expert in that area, but
I believe that our intent was to be consistent with the
current licensing basis of the plant.
MR. STROSNIDER: These are processed as license
amendments, changes to the technical specifications, and
that becomes part of the licensing basis of the plant.
It is a provision, but it is done on the licensing
basis. Does that answer your question?
MR. HIGGINS: Generally.
MR. STROSNIDER: I was just going to comment that
I wrote down two things that we need to get back to the
committee on: One was some additional description with
regard to the adjustment methodology used for taking the
leakage data and adjusting it for pressure and temperature.
And the other was, I think, a little more
background on the ten to the minus fifth criteria that came
from NUREG 0844. So we'll get back to you with those two
things.
DR. CATTON: Do we have somewhere, a figure that
shows the measurements of leak rate as a function of
voltage?
MR. KARWOSKI: That would be in the proprietary
handout. All the correlations of leak rate as a function of
voltage are in there.
This came out of a more detailed report which has
much more details with respect to, you know, crack
dimensions and whatnot, but this is an excerpt of the
database, so it shows the actual data plotted.
DR. KRESS: Jack, it was five times ten to the
minus two.
MR. STROSNIDER: Yes, you're right, 0844 was five
times ten to the minus two, and we adjusted it to ten to the
minus second, so we'll --
DR. KRESS: You said ten to the minus fifth.
MR. STROSNIDER: I'm sorry.
DR. KRESS: You confused me.
MR. STROSNIDER: Thank you for telling me.
CHAIRMAN POWERS: A modest difference. Let me
make sure I'm going to be able to derive everything I need
to know about this log logistic curve for the rest of the
members.
DR. CATTON: This looks a lot better than some.
DR. KRESS: If you have questions about that, you
should ask them now.
DR. CATTON: I guess I just basically understand
what the correlation is used for, and by the particular one
that was selected. I mean, it's an obscure one. To say,
that I don't know what a logistic is, is to overstate the
case a little bit, but it's not the first one that comes to
mind.
CHAIRMAN POWERS: Just to help you before you go
on, is this the data that you're talking about that has no
correlation?
MR. KARWOSKI: That's the 3/4 inch. There's a
much stronger correlation. There should have been another
handout that you had.
What I mentioned earlier, a tube either doesn't
leak or it leaks and so what we have here is probability of
leakage as a function of bobbin voltage. If you were to
look at the actual data, you would see tubes with certain
voltages that didn't leak and tubes with certain voltages
that did leak, and you have the actual data in the handout.
The proposal made by the industry was to fit a log
logistic curve through this data.
There is no theoretical basis for the log
logistic. There could be other curves that fit through the
data.
As part of NUREG 1477 the Staff looked at some of
these other correlations and what this plot has are some of
these other correlations.
The only thing I want to point out is you can see
that for this function you will have a higher probability of
leakage than these functions.
These curves criss-cross each other also, so which
curve will be conservative, which is a function of what the
voltage distribution is to some extent because depending on
what your distribution is you -- depending on which
probability of leakage curve you use, you'll get a different
answer, so as part of NUREG 1477 the Staff looked at various
correlations and various distributions.
I forget the exact numbers but if you have a
distribution centered around one volt you may conclude that
the COSHI is conservative with respect to the other models.
If you go to a higher voltage at 1.5 volts based
on some of these correlations criss-crossing you may
conclude that another model is conservative like the log
logistic or the log normal.
What the Staff then did was say, well, which
correlation should we use or which function should we use,
and what the Staff concluded was given the other
conservatisms in the model and the fact that each one of
these functions had roughly the same degree of fit to the
data that the log logistic was acceptable and that is why we
used the log logistic.
CHAIRMAN POWERS: I think you may be too strong
when you say there is no rationale for it, because log
logistic is an extreme value distribution, so what you have
got here is an on/off switch sort of phenomenon fitting in,
and so I think there's more justification for it than maybe
the normal distribution.
Now the question that comes to mind is that my
recollection of the data -- I have to look again to remind
me -- is if we take a value of, say, 4 volts, maybe 5, we
will find tubes that had 5 volt indications that leaked with
100 percent probability because they were tested.
MR. KARWOSKI: Right.
CHAIRMAN POWERS: Okay, and at 4 volts you would
find tubes that didn't leak with 100 percent probability
because they were tested, so now I am wondering why would I
want to use a continuous high entropy distribution like this
at all for this.
Why wouldn't I just come through and say, okay, my
minimum voltage at which I ever observed a leakage was 4.5
volts, say --
MR. KARWOSKI: The industry would probably like to
do that.
Actually, I think they proposed that, below a
certain voltage they didn't need to consider leakage, but
that isn't necessarily conservative, so we fit a continuous
distribution to it because you get into how much overlap of
the data do you have -- do you have enough datapoints to
sample the actual data?
Although there are a lot of datapoints in this
overlap region I don't think there's enough data.
CHAIRMAN POWERS: So what you really were trying
to do is put a non-zero probability for leakage on those
that did not observe leakage in that voltage range where
your -- all your datapoints said there was no leakage. You
wanted to put a non-zero probability there.
MR. KARWOSKI: Correct.
CHAIRMAN POWERS: And you did that --
MR. KARWOSKI: This is an imperfect correlation.
CHAIRMAN POWERS: And you did that at the expense
of putting a nonunity probability for -- in that voltage
range where you had datapoints that said that it would leak?
MR. KARWOSKI: Right, but on the other hand this
data overlaps. I think if you look at the data, I would
have to look, but there's points which don't leak at like
maybe 8 volts and I am making these numbers up, and things
that leak at 6 volts and vice versa so you are right.
But given the other conservatisms in the
methodology the Staff believed that that was an
acceptable --
CHAIRMAN POWERS: Well, I may be preaching to the
choir, but coming in and defending an action because there
are other nonconservatisms all on top of nine conservatisms
just emphasizes the fact that we end up not knowing what the
conservatism of the whole is.
DR. KRESS: The log logistic has a cutoff at the
upper end? It goes to one at some value?
CHAIRMAN POWERS: Well, I think it goes near one.
MR. KARWOSKI: It goes near one.
DR. KRESS: That's why they call that the
infinity?
CHAIRMAN POWERS: I think my recollection is --
DR. KRESS: I thought it went to a value that you
had to stick in there.
MR. KARWOSKI: There are functions like that.
This one doesn't -- where you specify at what voltage you
get a zero. There are functions like that and the industry
has used them in POD but they did not use that here.
CHAIRMAN POWERS: I mean I think my recollection
of it, it's very hard to invert because that is given a
probability of what is the value on the horizontal axis it
is hard to invert because it goes up so tightly --
DR. KRESS: Essentially one.
CHAIRMAN POWERS: That is my recollection on the
thing but I can't swear that it doesn't actually --
DR. KRESS: Does it really have that little
discontinues?
MR. KARWOSKI: It definitely does not have that.
MR. STROSNIDER: This is Jack Strosnider. I
wanted to make two comments.
One is just with regard to Dr. Powers' discussion
about putting a non-zero for these low voltages. Part of
the discussion that the task group had when we were
developing 1477 was the likelihood that you could have a low
voltage indication but that voltage might be coming from one
crack as opposed to a network, and the possibility that
there is an outlier, more or less, and I am not sure that
this completely addresses that issue, but there was some
discussion about should it ever really be zero, likelihood
of that sort of thing.
CHAIRMAN POWERS: I mean the statement was there
was no technical basis for it and in this discussion I found
at least two.
One is it's an extreme value, it's tradition, so
it's appropriate for these things, and the rationale was
putting on sort of probabilities for whatever reason down
there.
MR. STROSNIDER: The second comment I wanted to
make was come back, changing subjects on you here to the
correlation issue again, and Ted Sullivan of the Staff just
pointed out to me that there was actually more going on than
just adding data to the database, that in fact the industry
realized that they were not applying the P test correctly.
They were doing a two sided test and it should have been a
one sided test and they did the one sided. The correlation
popped out.
If you look at the original data you might
conclude it was there all the time then. I don't know, but
there was a little more to the story than what I gave you
before.
MR. CATTON: In looking at these figures, for the
three-quarter inch tubes it looks like there is a pretty
strong dependence on the voltage and it's rather weak for
the seven-eighth inch. Is there any explanation for that?
That's only an eighth of an inch difference.
MR. KARWOSKI: Right and they're scaled tubings
with respect to the tube.
I believe the industry would argue and I would
have to go back and look because it's been awhile since I
have looked at that database, the industry has proposed
excluding certain datapoints based on various reasons, and
they would argue that some of those leak rates are
inappropriate.
The Staff did not agree with all the exclusion
criteria that the industry wanted to apply to the data.
MR. CATTON: Which, the seven-eighths?
MR. KARWOSKI: To both. It actually applies to
both but I think the outliers were more in the seven-eighths
inch database, but I would have to look it up.
MR. CATTON: So when this process is exercised,
the one that you explained, probability, voltage
distribution and so forth, you then come to the break flows.
If it is a three-quarter inch tube they use this figure, the
figure 6 that is in this document? This one?
MR. KARWOSKI: They would use that.
MR. CATTON: And if it is seven-eighths they would
use this other one?
MR. KARWOSKI: Correct.
MR. CATTON: Thank you.
MR. STROSNIDER: I think it's empirical.
MR. CATTON: I can't understand the difference
really, but --
MR. KARWOSKI: There's also observed differences
in the burst pressure with respect to the one and two volts.
MR. CATTON: One has a thicker wall.
MR. KARWOSKI: One has a thicker wall but the
diameters are different so the ratios are essentially the
same.
MR. STROSNIDER: Actually there were some
theoretical analytic studies done back when this was being
developed to try to understand that difference, R over T
ratios, et cetera, and I don't think anybody could ever put
their finger on it.
I don't think we have a good answer.
MR. CATTON: Are there more generators for
three-quarter than seven-eighths?
MR. KARWOSKI: Actually more -- with respect to
the people who apply it, more are seven-eighths inch than
three-quarter inch, the Sequoyahs, the Beaver Valleys, the
Farleys, Diablo Canyon, Prairie Island -- they are all
seven-eighths.
I think Watts Bar and South Texas are the only two
that are right now three quarters.
MR. CATTON: Because you got some really low
values, very high voltage, and those are built into all the
averaging. That's not comforting.
MR. KARWOSKI: As Jack pointed out, I would like
to now talk about steam generators in general rather than
Generic Letter 9505, although I will pull in portions of
9505 in this discussion.
With respect to tube repair criteria that have
been approved, the two dominant ones are the 40 percent
depth base limit, which was developed 25 years ago and you
have the Generic Letter 9505, which is the voltage-based for
ODSCC at the support plate.
The crack growth assumption in the 40 percent tube
repair criteria we kind of discussed this morning, but the
growth rate assumption in there was that the NDE uncertainty
in the growth rate was somewhere on the order of 20 percent
and they assumed infinitely long degradation.
The crack growth rate for the alternate tube
repair criteria is either based on plant specific data if
you have enough datapoints or bounding generic data if you
do not have enough datapoints.
One of the questions that cam up is why don't we
use laboratory crack growth rate data. While there's many
factors that influence crack growth, there's operating
parameters like temperature. There's water chemistry, bulk
versus crevice, how well do you know the type of crevice
chemistry that you are having, the tube material affects
crack growth.
Many of these factors, and it is not intended to
be all inclusive, but many of these factors are not only
plant specific or steam generator. They can be in some
cases tube specific with respect to what is happening in the
crevice.
It is difficult to apply laboratory growth data to
the field because the assumptions made in the laboratory are
usually conservative to try to bound a variety of conditions
and they may or may not be representative of what is
happening in the steam generator.
For that reason, you know, in Generic Letter
95-05, it's a voltage-based approach. There really is no,
quote/unquote, crack growth; it's how much voltage
progression do I have over the course of the cycle.
This basically describes the methodology for
determining the growth that I previously described this
morning. And there are two growth rate calculations that
are performed:
One for the deterministic determination of the
structural repair limit of 5.5 volts, and that's basically
an average growth rate.
And then there's a growth rate distribution which
is used in the Monte Carlo analysis for determining the
conditional probability of burst.
It's in this where if you do not have enough
datapoints, that basically you'll use the bounding generic
database, rather than a plant-specific.
And based on the predictions of the end-of-cycle
voltage distribution, in general, the growth rate -- I'm
sorry. In the predictions of the end-of-cycle voltage
distribution, negative growth rates, as we discussed before,
are treated as negative in the Generic Letter 95-05
approach.
DR. BALLINGER: Is treated as zero, right?
MR. KARWOSKI: Is treated as zero.
DR. BALLINGER: You said they were treated as
negative?
MR. KARWOSKI: Negative growth rates are treated
as zero.
So, the real question is, with these growth rate
distributions -- and really the only growth rate
distributions that the staff uses is for Generic Letter
95-05.
As I pointed out earlier, those predictions are
usually pretty accurate. I'm sorry, the predictions of the
end-of-cycle voltage distributions tend to be conservative,
as I pointed earlier, and licensees are required to evaluate
their inspection results as a result of NEI 9706.
DR. SIEBER: If you were using the tech spec, the
old time tech spec values for maximum flaw depth of 40
percent through-wall, that has built into it, a ten-percent
growth, roughly.
MR. KARWOSKI: Roughly.
DR. SIEBER: How do the actual flaw growth data
compare to the ten percent, if it's built into the 40
percent through-wall?
MR. KARWOSKI: There will be tubes that exceed it
and tubes that are less than that. In general, for the
wastage and wall thinning, it would be plant-specific on
what growth rates they observed.
It's only until NEI 9706 that licensees now start
doing those condition monitoring and operational assessments
where they start doing more detailed assessments of what the
conditions will be at the end of the cycle.
DR. SIEBER: Now, it would be interesting to know
at some time, whether the original 40 percent through-wall
was conservative or not.
MR. KARWOSKI: With respect to when you include
NDE uncertainty and growth?
DR. SIEBER: That's right.
DR. BALLINGER: Well, when you look at it -- I've
been thinking about that, and if you take a look at all of
the steam generator tube ruptures that we've had, it's
pretty much been independent of whether anybody has been
applying the 40-percent criteria or not.
DR. SIEBER: That's right.
MR. MURPHY: Ken, if I might add one thing, this
is Emmett Murphy.
Condition monitoring programs have been conducted
by licensees routinely now for several years. In general,
these condition monitoring programs are successful in
demonstrating that all tubes have adequate margin at the end
of the cycle.
So, this early experience indicates that a
situation where implementation of the 40-percent plugging
limit, in general, that approach does ensure that adequate
margins are maintained at end of cycle; not always, but in
the vast majority of the cases.
CHAIRMAN POWERS: I think what you said is
absolutely true, but it looks to me like it's close in some
cases. I mean, we're getting close to the ten to the minus
two acceptance probability for burst in the Farley example.
And it looks like in the projection of the 15th
cycle, you know, they were close. There's some of Tom's
virtual character to that 15th cycle, I'll admit.
I mean, I guess the issue is, is there -- are the
acceptance limits set sufficiently high that getting close
is not a source of concern? And correct me if I'm wrong,
Jack, but in the original discussion, that acceptance limit
was really five times ten to the minus two?
MR. STROSNIDER: If you go back, again, talking
about Generic Letter 95-05, and the original assumption in
NUREG 0844 was five times ten to the minus second.
And we reduced that by a factor of five to account
for the fact that there could be more than one degradation
mechanism.
The thing I'd ask you to think about, though, is
putting the voltage-based criteria aside, criteria that are
in the licensing basis that the plants need to meet are
three-delta-P on normal operating pressure, and 1.4 on main
steam line break.
And what's being done now in terms of the
condition monitoring is not only eddy current testing, but
in situ pressure testing, all right? And I think what
Emmett was indicating is that in the majority of case where
they look at what they think is the worst defect in the
steam generator and they do this in situ testing, it
satisfies those factors of safety.
So, moving away from the voltage-based, there are
other criteria that come into play.
CHAIRMAN POWERS: I guess the nagging concern
here is that there are a lot of point examples, and I don't
have enough collection data together to get out of the --
effect, you know, enough cases to persuade me that this
isn't just a quick of nature, and that tomorrow we have one
that's an egregious example.
MR. STROSNIDER: Well, I guess I can offer two --
the only response that I think I can give is, one, we'll
continue to collect data, all right, and hopefully collect
more confidence in that regard until the generators are
replaced and nobody's dealing with this one any more.
CHAIRMAN POWERS: The real answer.
MR. STROSNIDER: Yes, but I'd also come back to
the point that I made earlier, that there is always the
potential for some new form of degradation or for some
change in degradation mechanism in terms of growth rates.
I would point out that the industry has moved to
more stringent controls, for example, in water chemistry.
There is more consistency there. They are looking now in
terms of corrective actions.
People, on occasion, will lower operating
temperatures, so in that sense, things are being done that
can be done to try to make things a little more predictable.
However, you can always get to a point where you
reach the incubation time for a new type of degradation and
it shows up.
CHAIRMAN POWERS: We have the famous 13.7 or 14.6
that seems to come about because of something unanticipated.
MR. STROSNIDER: Right, and if you look at the
tube rupture list, you will see that there are a number on
there that were caused by loose parts.
Just a few other observations there: One is that
with regard to performance criteria -- and we weren't
planning on getting into 9706 and the operational
assessments and that sort of thing a whole lot, but we
really did take a look at trying to establish performance
criteria that left sufficient margins such that even if they
were violated, it didn't represent the end of the world.
Similarly, I think -- and as I indicated, we'll
have to get back to you with some more detail on the 0844
evaluation, but my recollection is that from a -- and let me
characterize it as a risk-based point of view -- that you
actually could have driven that conditional failure
probability higher, an still come up with an overall
acceptable risk, given the frequency argument.
And, in fact, we had that discussion with the
industry where they wanted to use a much higher acceptance
criteria, and we felt that in terms of maintaining margins
and defense-in-depth, that we needed a lower number.
CHAIRMAN POWERS: Have you given thought to the
issues that accompany --
MR. STROSNIDER: I'm sorry, I can't hear you.
CHAIRMAN POWERS: Have you given thought to what
happens as we go to higher levels of burnup or higher
boration of the water or higher power, operating power
levels?
MR. STROSNIDER: With regard to power up-rates, a
look at the steam generators is, in fact, part of our
review.
I'm not aware that any significant issues have
come up with regard to those reviews. I can't really go
into a whole lot more detail than that.
CHAIRMAN POWERS: Does boration level cause any
headaches?
MR. STROSNIDER: Not that I'm aware of.
MR. HOLOHAN: I think licensees, at least to date,
have not been going to additional boration as a part of
longer cycles of power uprates. They're using burnable
poisons, and that goes to the poison concentrations are
controlled by things like moderator temperature
coefficients.
Now, if we were to relax our limits on things like
moderator temperature coefficients, then I think the
chemistry might -- I think the industry would like to use
more soluble boron and less burnable, but we haven't done
that to date.
MR. KARWOSKI: When I first started off, I said
there were three issues that I was going to talk about. The
first one was steam generator regulatory framework and
operating experience, and I did that this morning.
The second one was Generic Letter 95-05, its
technical basis, and also included a discussion of growth
rates, and I just completed that.
The third topic was NDE capabilities with respect
to detecting and sizing flaws, and that's what I'd like to
discuss now.
The primary means for inspecting the steam
generators tubes is eddy current testing. There are a
variety of different probes that are used during the
inspection of a steam generator tube, and I'm talking now in
generic terms, not Generic Letter 95-05, specifically.
The bobbin probe is the tool that's frequently
used just for screening the tubes for defects. It's
relatively fast, can do 24 to 48 inches a second, but it's
relatively insensitive to circumferential degradation.
Okay, it's also poor at characterizing
degradation. As a result, there's an alternate probe that
is used by licensees, and those are frequently referred to
as rotating probes, rotating pancake coil probes.
There are various types of coils that go on the
probe. It can be a pancake coil that is sensitive to axial
and circumferential flaws; a plus-point coil which is
sensitive to both also; and axially-wound coil that is only
sensitive to circumferential flaws; and a
circumferentially-wound coil that is sensitive to axial
flaws.
DR. CATTON: Just a quick question: Why is it
that they chose the bobbin to do these leak rate
correlations, when it's the worst measurement speed.
MR. KARWOSKI: It's 100-percent inspection. It's
sensitive to the axial type of degradation that occurs
there.
DR. CATTON: But I thought that before they pulled
the tube, they went in with the rotating probe to be sure
that whatever they saw was right. So don't they have the
rotating?
MR. KARWOSKI: They would have the rotating probe
data for the pulled tubes, because they --
DR. CATTON: Which is the database.
MR. KARWOSKI: Yes, but they don't do rotating
probe inspections at every intersection in the steam
generator. And you need something -- as a result, you need
something to correlate those intersections that have those
bobbin indications to something, and you're not doing
rotating inspections at every elevation.
DR. CATTON: But if you're trying to develop a
correlation, and you've got such huge data scatter, why do
you use the worst measurement for your correlation?
MR. KARWOSKI: I don't know if it's the worst. I
don't know how the correlation of both RPC voltages would be
compared to like burst pressure or to leakage.
DR. SIEBER: Maybe I can answer that. The bobbin
coil probe moves at about two feet a second.
DR. CATTON: Oh, I see this up here.
DR. SIEBER: So it flies through the steam
generator and you can examine two or three thousand tubes in
seven days, eight days.
A rotating pancake coil goes a half an inch to six
inches a second, and it just takes forever, and so it's a
matter of what can I use to keep my outage down to the
minimum, still do the inspection, and get reliable data.
DR. CATTON: I understand this, but maybe I'm
missing something in the process. You do the bobbin coil,
and then where it looks like you might have a problem, you
check it with the rotating probe before you pull the tube,
or do you pull the tube?
MR. KARWOSKI: Okay, I have to answer this in two
parts: When licensees pull tubes, they will stick a variety
of probes through there. So, for the tube-pull database,
yes, you would probably -- you probably have some data with
respect to the rotating pancake coil voltage at specific
locations on that -- along that crack.
The bobbin coil integrates around the
circumference, okay?
DR. CATTON: I understand.
MR. KARWOSKI: So for the pulled-tube database, I
am sure that there is RPC data for the vast majority of the
pulled-tube database.
When you do the inspection, though, you don't
RPC-evaluate every intersection. As a result, you would
have nothing to compare to your correlation.
DR. CATTON: You don't RPC?
MR. KARWOSKI: RPC, rotating pancake coil probe.
You don't do RPC probe inspections at every intersection.
DR. CATTON: You do it at the intersections that
the bobbin coil told you might be a problem.
MR. KARWOSKI: Only if the voltage is above one
volt for 3/4 inch tubes, and two volts for 7/8ths inch
tubes, so you don't have that inspection data.
DR. CATTON: So above one volt and above two
volts, you could check bobbin, RPC leak rate, and then you
could generate a nice correlation.
MR. STROSNIDER: This is Jack Strosnider, and I
want to point out one other thing here. It's not clear in
my mind that the RPC is going to be easier to correlate to.
As Ken pointed out, the RPC is a much smaller
probe, and it examines -- I mean, you get an actual profile
around the tube, so you're going to have to say, you know,
which part of that profile do I want to make the correlation
with? The peak? The average, or something else.
The bobbin probe is sort of an integral look at
that intersection. And so at first you'd have to decide
that it may be possible that you could go through there and
pick something off of there to correlate, but I wouldn't
assume that I could necessarily get a better correlation.
DR. CATTON: But if you did -- but I might be able
to explain why I got three decades variation in the data,
because I would know more about what it is I'm looking at.
CHAIRMAN POWERS: You don't get a single number
out of an RPC; you get a flattened out geography, 3D
profile.
DR. CATTON: That might help the mechanics guy
figure out why the hell the thing is leaking.
MR. STROSNIDER: And as I mentioned earlier,
though, if you go run leak tests, not on intergranular
stress corrosion, not on this network, ODSEC type cracks,
but on single stress corrosion cracks, you will get the same
sort of distributions in variability in leakage.
CHAIRMAN POWERS: If you've got an absolutely
perfect correlation with the RPC, you would still have to
have your bobbin coil distribution.
MR. KARWOSKI: Or we would have to change the
Generic Letter and require licensees to inspect with a
rotating pancake coil at every support plate.
CHAIRMAN POWERS: Yes, you could do that, but
given that you didn't do that, you'd still need a bobbin,
and it would still look just like it does.
MR. KARWOSKI: Right, you would need something to
correlate.
CHAIRMAN POWERS: They've got to account for all
of the indications that they find in carrying out the
process, not just those that are over some voltage limit, be
it one volt or two volts.
MR. KARWOSKI: That's right.
CHAIRMAN POWERS: You have to invert those
distributions in some way to end up with a predicted leakage
rate.
MR. KARWOSKI: That's right. Another probe is
used. It is a Cecco probe. It's a transmit, receive and a
ray-type probe. It is medium speed, around 12 inches per
second. It's sensitive to axial circumferential flaws.
Some utilities choose to use the Cecco probe
instead of the bobbin and rotating. Other utilities feel
more comfortable using the bobbin and rotating. It is up to
the utility. The tech specs do not dictate what probes to
use.
The other thing that I put on this slide is that
ultrasonic testing is also sometimes used to inspect steam
generator tubes.
Usually it's as a supplemental technique to help
characterize degradation.
CHAIRMAN POWERS: Before you launch into that, I'd
just raise a question that the former Commissioner Rogers
raised on several occasions.
Is there a growing technology in this area? Are
people trying other kinds of technologies to do better on
these things?
MR. KARWOSKI: The technology is evolving. A
couple of the slides I have, there's different algorithms
used.
There's new probes that are coming out. A lot of
them are still based on eddy current technology but there
are advances being made and you will hear about some of the
work being done at one of our contractors.
The probes used to inspect steam generator tubes
have changed over time. Originally in the 1970s they used
the single frequency bobbin coil and I think somebody
described it before where they had a oscilloscopes where
they were analyzing the data.
That was good for general wall loss type of
degradation mechanisms, not so good for stress corrosion
cracking.
In the last late '70s and the '80s, multiple
frequency bobbin coil techniques started to be developed
along with rotating pancake coils probes. In the early
'70s there were no rotating probes.
The multiple frequencies allowed mixing out some
of the unwanted signals and it allowed you to focus at
different parts of the tube while just pulling the probe
through once rather than several times.
The rotating pancake coil probe, as I said
earlier, was better at detecting and characterizing stress
corrosion cracking. It is better in geometry changes. It
is sensitive to circumferential flaws. It was initially
used primarily at the expansion transitions.
Widespread use of the rotating pancake coil probe
began in the late '80s, early '90s. People, licensees,
really started to inspect their expansion transitions.
The plus point coil, which you probably heard
about, merged in the mid-1990s. Its first major application
was at Maine Yankee, where they had problems with
circumferential cracking at the expansion transition and new
probes and data analysis software and techniques continue to
be developed.
So what drove these improvements in technology?
Well, both economics and regulatory concerns. If you
remember from this morning I showed a plot of the forced
outages as a result of leakage. Back in the '70s and early
'80s there was a number of outages. Outages are costly.
What caused those leakage outages? Well, it can
be a variety of factors.
It could be the technique capability -- how good
was that single frequency bobbin coil at really inspecting
the tubes for the degradation mechanism of interest?
Could have been analysts' reliability. How well
were those analysts trained on those techniques? A lot of
those techniques were evolving at the time.
Could have been high growth rates.
It could have been any combination of those three.
In the mid-'80s the Office of Research had Pacific
Northwest Laboratory analyze the removed steam generator
from Surry. They shipped the steam generator from Surry up
to PNL -- Hanford? -- and they did a bunch of analysis.
They did leak testing, burst testing. They developed
various correlations.
Under that contract they also determined some
probability detections. The dominant degradation mechanism
or one of the prevalent was wastage. However, as part of
that program they did a mini-round robin of laboratory
samples, not from the Surry steam generator -- laboratory
grown samples, and determined the probability of detection.
The probability of detection varied from team to
team and I don't recall the exact ratios but some of them
were pretty poor -- .3 probability of detection -- to
larger, maybe .8 POD independent of depth.
The average turned out to be about .6, and that is
where we got the value for the generic letter 9505
correlation, so we used a POD of .6 based on techniques
available in the mid to late '80s and we applied it to the
Generic letter 9505 approach.
Much has happened since then. In the '80s and
early '90s the industry started developing ISI guidelines.
They needed to be able to detect this degradation sooner.
They didn't want to have forced outages. They wanted more
reliable inspections.
These guidelines started to specify criteria for
the probability of detection. Some of the versions in the
early '90s said we could have an 80 percent probability of
detection at 90 percent confidence and there's various
criteria with respect to what are the depth distributions of
the degradation.
The only thing I just wanted to point out is that
the industry guidelines really started evolving in the late
'80s and early '90s.
They also developed a qualified data analysts
program, a rigorous training program, with the analysis
qualified to detect specific types of degradation.
The other issue that was starting to emerge in the
early '90s was generic versus plant specific qualification
of techniques. The role of plant specific factors in the
detection of flaws became a concern. Yes?
DR. SIEBER: About two hours ago you were talking
about the NDE uncertainty distributions.
MR. KARWOSKI: Yes.
DR. SIEBER: And you said it was made up of two
elements, analysts' variability -- that was about 10
percent? -- and physical variability, which I presume
includes systematic and random calibration errors, probe
wear --
MR. KARWOSKI: It was predominantly probe wear.
DR. SIEBER: Okay, because other factors were in
there?
MR. KARWOSKI: It was specifically probe wear.
They shaved a probe and analyzed what the change in voltage
response was as a result of physically wearing down a probe.
DR. SIEBER: How did they accommodate things like
calibration errors and random errors that occur just in the
process of any kind of sampling technique?
MR. KARWOSKI: All of those types of errors will
be captured and all of the correlations, because all those
pulled tube datas will have all those different errors
promulgated through it.
DR. SIEBER: How did they come up with the number
for analysts' variability?
MR. KARWOSKI: Analyst variability was by having a
group of analysts evaluate several different -- well,
hundreds of different indications and determine what was the
difference between the voltage calls between the
different --
DR. SIEBER: Who is smart and who is not, right,
essentially, so there is actually a basis for these
distributions that go into that for every element of it?
Do you have a list of all the elements, what the
distribution looks like?
MR. KARWOSKI: For the two distributions, there
are only two distributions --
DR. SIEBER: Only two.
MR. KARWOSKI: -- that the probe wear --
DR. SIEBER: And the analysts.
MR. KARWOSKI: -- and the analysts' variability,
and we do have that data. We can provide it to the
committee.
DR. SIEBER: If it is not too lengthy --
MR. KARWOSKI: No, it is actually --
DR. SIEBER: Two pages. Thank you.
DR. BONACA: You said before that the POD, the
licensees contend that the POD depended on the entity of the
defect, that for a bigger signal you would have -- do you
have any inputs on distributions from the licensees?
MR. KARWOSKI: Yes. I will show you several of
those.
DR. BONACA: Okay.
MR. KARWOSKI: In the mid to late '90s the
industry and the NRC developed additional guidelines which
addressed various factors like plant specific considerations
and the concern of what is the probability of detection. Is
the technique capability or is it the entire system?
There is currently an evaluation of steam
generator mockup samples being performed by Argonne National
Laboratory as a result of a contract with the NRC. At the
very end of this presentation I am going to ask Dr. Muscara
to present some of that work and their results with respect
to detecting and sizing flaws.
There are a number of factors that affect
detection of flaws. There's equipment and technique
variables. There are -- the analyst plays a role in
detection. There's also plant specific considerations.
With respect to the essential variables, there are
equipment variables -- what equipment do I use to get the
data, what type of acquisition system, what type of probe am
I using, what type of cable, is there enough shielding in
that cable to prevent noise signals from being picked up.
There's technique variables. What are the
frequencies, the dry voltage calibration method, how much
data am I obtaining, the digitizing rate, how do I scan the
tube -- what is the direction, do I gather the data on the
push or pull?
There's analysis variables that also affect
detection, including what the data review requirements are,
the algorithms used in the software, and the calibration.
Of course, analysts' reliability plays a role in
detection and there is also plant specific considerations.
There's the role of deposits. Are those deposits
conducting, nonconducting, ferromagnetic. How do those
deposits affect the signals? Dents and geometry changes
affect your ability to detect degradation. Support
structures. Do I have other interfering signals that are
coming in. The crack orientation, is it axial,
circumferential? Is it on the ID or the OD of the tube?
Also there is noise.
CHAIRMAN POWERS: You have noise, electro:
tube -- is it tube noise?
MR. KARWOSKI: Tube noise is basically the surface
of the tube is irregular. Also tube noise can come from
deposits. They kind of overlap. Noise comes from a variety
of sources.
Okay. The industry qualifies specific probes,
specific sized probes for specific degradation mechanisms
under specific circumstances at specific locations, so when
we say what techniques are qualified for detecting, it
depends on what you are looking for and how you are looking
for it.
Each one of these techniques has a list of
essential variables, so what does this mean?
Basically a given size plus point coil is
qualified to detect circumferential primary water stress
corrosion cracking at a specific frequency in dented
locations. The industry has a dataset which they believe
demonstrates that they have an 80 percent probability of
detection at 90 percent confidence.
The industry has lists of essential variables that
go with each one of these techniques. I need to keep this
probe or I need to acquire the data in this fashion and the
dataset that supports qualification has this much noise.
The industry has an extensive program with respect to the
ability to detect degradation.
There are a number of issues that have been raised
with respect to what is the probability of detecting flaws
under certain circumstances, and I have listed some of them
here.
One of them was axial versus circumferential
cracks. How does your ability to detect degradation depend
on the orientation?
Well, the best way I can answer it is it depends
on the probe. A bobbin probe will not reliably detect
circumferential degradation. Unless that circumferential
crack has opened up a lot axially it won't seal.
The bobbin probe also has weaknesses at detecting
axial degradation under certain circumstances, namely in
dents, severely dented tubes, U-bends, and expansion
transitions.
The pancake and plus point coil, on the other
hand, are qualified for detecting those cracks which the
bobbin coil would not be qualified for. For example, axial
degradation at those locations or circumferential --
CHAIRMAN POWERS: Initial screening done with a
bobbin doesn't detect circumferential cracks reliably, that
is what you are saying?
MR. KARWOSKI: Right.
CHAIRMAN POWERS: And are those, is the rationale
then for using the bobbin coil for the initial examination
then that circumferential cracks are sufficiently rare that
missing them if of little consequence?
MR. KARWOSKI: Licensees in general do not
normally just use the bobbin coil. The industry is aware
through numerous generic communications that the potential
for circumferential cracking exists at various locations,
namely at the expansion transition, at dents and U-bends,
and in sleeves, and as a result consistent with Appendix B
they use techniques that are qualified for detection.
If you look at what licensees do, they would, in
those locations they would do some type of sampling program
to ensure that they are detecting the forms of degradations
that those tubes are susceptible to.
The Staff has issued generic communications
highlighting the weaknesses.
CHAIRMAN POWERS: So I understand this better,
talk about circumferential cracks in the free span area.
Are those sufficiently rare that you think there is no need
to go around looking for that?
MR. KARWOSKI: There have been no circumferential
cracks in the free span area.
CHAIRMAN POWERS: That's why I picked the example.
MR. KARWOSKI: The answer is yes.
You would have to look at are there any stresses
sufficient to induce that type of cracking in the free span,
and in general the likelihood of that is very small and
so -- and that supports operating experience where none have
been identified either through leakage or through
inspection.
The bobbin coil I did say is relatively
insensitive, unless those cracks open up. If it were
happening in the free span in a nondented area, you would
have either, you know -- the cracks would eventually start
growing and you would either detect it by bobbin or through
leakage. We haven't observed that today.
The other issue is, how well can I detect with
respect to isolated cracks versus cracks in clusters? The
comment that I make there is that, in general, the more
material you use, the less ligament paths that you have, the
easier it is to detect that type of degradation, but
characterization of the flaw is much more difficult.
DR. POWERS: Is there a well recognized
description of the width of cracks?
MR. KARWOSKI: Of the what?
DR. POWERS: Width of cracks. I mean a very, very
fine crack, different from one that is opened up a ways.
MR. KARWOSKI: I am not sure I understand what you
are asking.
DR. POWERS: Well, it is this volume, missing
volume. The more missing volume, the easier it is to detect
clusters. What I am thinking is, suppose I have an isolated
crack and have a crack that is really hairline wide and has
lots of cross ligaments throughout its length, or I can have
another isolated crack that actually has some canyon, open,
it is a canyon, I can put something down into it.
MR. KARWOSKI: Right.
DR. POWERS: The two are very different, and I am
wondering how you -- I am having a hard time describing
these cracks. Is there some measure or something like that
that does a better job than I am on describing these two
different types of cracks?
MR. KARWOSKI: In terms of NDE, no. But you could
describe what the crack opening area is, but in terms of
NDE, you know, the only thing you would notice is that you
get a much larger signal from an EDM notch than from a
crack. I mean the wide open one is what I am thinking of.
DR. POWERS: What I am thinking of is there are
going to be some cracks, I mean I can imagine a crack,
whether one actually ever existed, and which, across the
length of it, there are ligaments and metal that are still
there.
MR. KARWOSKI: Right.
DR. POWERS: That might be very different than one
in which there were no ligaments.
MR. KARWOSKI: Absolutely.
DR. CATTON: And then if you fill those spaces
with some kind of crystals of something or other, it would
be completely different again.
DR. POWERS: Or if it was amorphous sludge, it
would even be worse.
MR. STROSNIDER: This is Jack Strosnider. That is
certainly true, and that is just describing physics and the
reality, what is out there.
I guess what I think is really relevant here,
though, and we have had a lot of discussions with the
industry, and if you look at our Draft Reg. Guide 1.174 that
was put out for public comment, when you talk about the
qualification process, it is very important to use what you
might characterize as prototypical cracks. And you give a
perfect example where the use of EDM notches may not give a
good qualification demonstration.
And the industry has made some progress in that
area. We continue to push on it because it is very
important. And some of what we are discussing in the 97-06
framework is, you know, the need to use cracks that have
signals that are representative of what is in the field,
whether they be autoclave or pulled tubes. There are some
EDM notches used in the qualification database, but, you
know, we push to minimize that. And I don't know if that is
part of what you were thinking about, but it is an important
aspect of recognizing the difference.
MR. KARWOSKI: Some other issues they came up
with, how is my ability to detect cracks that are plugged?
Once again, this issue depends on the nature of the
deposits. In general, deposits on the tubes are on the
outside surface of the tube. They are not -- these stress
corrosion cracks tend to be very tight.
But it would depend on the nature of the deposits.
If they are conducting, it is going to lower the eddy
current response, making it more difficult to detect.
Ferromagnetic deposits, you know, will increase the
response, but it will mask the flaw. If the deposits are
not conducting and no effect on tube noise, you probably
have similar detection probability.
The other thing is, what is my ability to detect
cracks relative to location with respect to tube support
plates or bends? In general, the more geometry changes that
you have, the harder it is to detect a flaw. However, there
are techniques that are qualified for giving locations in
order to provide detectability.
Operation speed was also -- how does that affect
your ability to detect degradation? Utilities have done
various assessments of this. The speed of the bobbin coil
has gradually gone up as the ability of the hardware to
process this data has gone up. Typically, the run the line
tests where they change the speed and determine whether or
not they would have detected the same tubes, you know,
regardless of the speed.
What is interesting to note, at first you would
think the higher the speed, the more difficult it is. At
one of the plants, they increased the speed, I believe the
ratio was 12 to 24 inches per second, but what they found
out, that increasing the speed actually reduced the noise
that they were observing and the detection was comparable.
That is not to say, though, well, let's go up to
the next speed and it is going to be better. There is a
problem with too fast a speed and that is because if you
have any geometry changes, which you do, in several
locations in the steam generator, that probe can jump,
resulting in noisy data, missed data and cause problems. In
addition, too fast a speed, because of the software that
processes this data, you can have some frequency effects if
you don't have proper compensating software.
The next part of the presentation shows you
probably detection curves for various degradation types.
These are all industry data. The staff has issues with
them. I will point some of these issues out as we go along,
but the committee indicated it wanted to know how we can do
with respect to detection. I am going to present some
industry data.
Let's start with wear, general wall loss type of
thing. In general, detection is pretty good. What this has
on this plot is the fraction detected as a result of the
throughwall extent of wear indications, under, you know,
given frequency indications. At face value, and this is the
industry data, from December '93, at face value, you might
say, well, I can detect 100 percent of flaws affected by
wear. The staff isn't saying that, but that is what this
data may tell you.
And at the end of presenting all these probability
of detection curves, I will go over some of the issues that
we have with respect to this data. But, in general, the
staff expects that, you know, sizing or detection of general
wall loss indications is pretty good.
DR. POWERS: What would it be at 99 C/L?
MR. KARWOSKI: What that is is what is the
probability of detection at 90 percent confidence. So even
though you detected 100 percent, you know, you only had 13
samples and there it is. Remember, the acceptance criteria
that I believe the industry is still using is 80 percent
probability of detection and 90 percent confidence.
Okay. This next plot is for plus point coil
detection in a sludge pile region. Once again, in this case
they fitted a curve to the data, you have probably a
detection as a function of maximum depth. You get into the
same issues, why a logistic fit versus other fits.
Basically, what this says is we have got somewhat of a high
probability of detecting larger flaws, a lower probability
of detecting smaller flaws. That is the sludge pile.
This is also for sludge pile. That first graph
was maximum depth. This one has several different curves as
a result of maximum depth and average depth. And as you
might expect, the probability of detection for the maximum
depth is less than the average depth, and that is because
you have a larger flaw when your average depth is greater
than just the peak. But, in general, for the average depth,
the probability of detection is larger.
The dots indicate the bobbin coil probability of
detection. This data, once again, industry data, indicates
that the probability of detection with the bobbin coil is
better than the plus point for these low voltage
indications. To what extent that is an artifact of the
distributions they chose, I don't know. I am just
presenting what they submitted.
This next graph is for --
DR. SIEBER: A question.
MR. KARWOSKI: Yes.
DR. SIEBER: The plus point has a whole bunch of
different coils, some of which look like bobbin coils and
some of which are rotating.
MR. KARWOSKI: The plus point coil is a specific
coil that will fit on a rotating probe.
DR. SIEBER: Okay.
MR. KARWOSKI: The plus point coil has an axial
coil and a circumferential coil.
DR. SIEBER: Okay. But you can --
MR. KARWOSKI: Okay. Wound together.
DR. SIEBER: -- mathematically play with it.
MR. KARWOSKI: Right. And one of the advantages,
or the advantage of the plus point coil, theoretically, is
that you have some general disturbance area, those two
coils, the signal from those two coils will cancel out.
That can also be a disadvantage of the crack if a crack is
perfectly, you know, symmetric and hits both coils at the
same time. But that is not -- that is theoretical general,
those types of flaws don't exist.
Here is a plot from December '93 of probability of
detection for outside diameter stress corrosion cracking at
the support plates. Staff used a value of .6 for the
Generic Letter 95-05 methodology. What the industry
presented in December '93 is basically for everything over
40 percent throughwall, we were detecting. The issues had
issues with respect to that, and that is why we used the .6.
But here is what the industry presented back in December
'93.
PWSCC occurs at dented tube support plate
elevations. Here is a plot of the fraction detected versus
maximum depth. What I will point out on here is this has
the fraction detected for the plus point coil, the bobbin
and the Cecco.
Just to point out some of the issues that you have
with limited data sets, you would conclude here that your
probability of detection may be higher in the 0 to 10
percent range than in the 10 to 20. That is an artifact of
the amount of data and shows some of the difficulties that
the staff has with respect to when somebody says, what is
the probability of detection? It all depends on the data
that supports the correlations.
The other thing to note is, you know, why would
you have a lower probability of detection at the higher
depths? And, generally, you would think the deeper the
degradation, you are better your probability of detection.
But, nonetheless, this basically shows, you know, reaching
towards 100 percent when the maximum depth hits around 50
percent for PWSCC at dents inspected with the plus point.
Once again, this is all industry data.
Here is another example for PWSCC at dented tube
support plate intersections. Some of the data comes from
that previous graph. I put this up because it compares the
bobbin and plus point, and this is a result of fitting a
specific function recommended by the industry through the
data. And what I wanted to point out here is here a
function, the bobbin POD is a function of depth, here is the
plus point, the two cross. In general, that is probably a
result of either the curve fit or the limited amount of
data, or the fact that, you know, you missed one indication
with the plus point and it results in a worse probability of
detection.
In general, the industry and the staff believe
that the plus point is probably better than the bobbin in
that area.
So how does the staff use the probability of
detection? Usually, probability of detection is only used
in Generic Letter 95-05 tube repair criteria. And in that
analysis, as I pointed out this morning, that we used .6,
and that is based on the roundrobin analysis of the
laboratory stress corrosion cracking samples that was done
in the late '80s.
As I pointed out this morning, that is used not
only to account for missed indications, but also indications
that may initiate during the course of a cycle.
And here I have a rhetorical question, is .6 the
correct value? I have showed you various data the industry
would argue for deeper flaws were better than .6, and there
may be some more data forthcoming from some of the Argonne
roundrobin tests that may support that, I don't know, but I
will say that the .6 is conservative and accounts for other
things than just missed indications, and that is based on
the operational assessments.
MR. HIGGINS: Do you know the breakup, breakdown
between missed and the ones that would initiate during the
cycle? And have you looked at end of cycle, beginning of
cycle test results for the last few years on 95-05 to see if
that number of those that initiate during the cycle is
reasonable with the actual data?
MR. KARWOSKI: The staff has looked at the 90 day
reports, and the industry has, with respect to, you know,
did I detect something in the prior cycle? The industry has
actually proposed an alternative way for POD where they look
-- and I don't recall this methodology in detail, the staff
hasn't approved it, but, basically, they do look back and
say, was there a signal there? And in some cases they
missed it, in which case that would be a missed indication.
In some cases, they would say, well, analysis wouldn't have
called that, so that would be a new indication.
I don't know if anybody has ever done a
comprehensive assessment of, you know, how many indications
are new versus how many indications were missed. I know the
industry has proposed an alternate probability of detection
model based somewhat on that type of approach, but, to date,
the staff hasn't accepted that.
MR. STROSNIDER: I would just add to that, it
might be difficult to separate the two, although, perhaps
with a hindsight review, when you detect something in this
outage, if you go back to the prior outage with some
knowledge, maybe you could say that it was there. But I
think the important think is you asked, is it being
benchmarked, and the sort of histograms, if you will, of
voltage versus number, you know, that is one of the reasons
we look at that, and one of the reasons we concluded that
the .6 does a reasonably good job of accounting for those
that are missed in inspection and new indications.
If we saw that those distributions were radically
different, one thing you could modify is that .6 factor and
try to bring them back into alignment.
MR. LONG: This is Steve Long. Let me add one
thing. By the time these things get to me it is usually
because they are doing some sort of risk-informed license
amendment request, so I guess I probably see the problem
ones more than anything else.
We had a problem with Arkansas Unit 2 this past
spring and their proposal to not perform another mid-cycle
inspection where they would have to take two inspections in
the middle of a fuel cycle. And we had a problem with
reconciling what they were projecting as the flaw
distribution of the tubes during the remainder of the fuel
cycle with what they had found in the last inspection and
what they had projected to find in the last inspection.
And it comes down to a question of a combination
of the growth rate and the POD in the previous inspection,
is your growth rate right or your POD right?
Arkansas was insisting that the things that they
missed in their inspection before last, they could find when
they did a lookback they were missed, and that their growth
rates were consistent once they found those things in the
lookback, that the growth from what they probably should
have seen in the inspection before was what they would
expect up from the last inspection.
So we asked them, then what does that say about
your POD? And I believe the POD was about .4, which was one
of the reasons we couldn't probably work with that, because
some of the larger flaws that we thought they would need to
be able to detect, there would need to be performance --
screening criteria. They really didn't have a very good
probability of detecting those in the inspection before
last, and we couldn't credit any change in the inspection
process for improving that over the last inspection.
So there are problems occasionally.
DR. HOPENFELD: This subject was discussed
yesterday and I pointed out that there is a possibility of
differences between running a test in a laboratory POD,
determination in a laboratory or at PNNL, versus in real
life, and I think this is a very good example, except that
this is only one.
And one of the questions that we didn't get to
discuss yesterday, this pertained to this very item, that we
need more information from plants to determine what that POD
is. Keeping this in perspective, I don't think it is proper
to say that what we are using is a conservative number. We
don't really know it is conservative.
DR. POWERS: Let me ask how much longer you have
to go on this presentation?
MR. KARWOSKI: A half hour.
DR. POWERS: Why don't we take a 15 minute break
now. We will recess until five of the hour.
[Recess.]
DR. POWERS: Let's go back in session. Pardon me
for interrupting you yet again, but I firmly ascribe to the
belief that the mind will ultimately absorb what the body
can endure.
MR. KARWOSKI: I am aware of that.
DR. CATTON: You know where the limits are, so you
are going to push them.
SPEAKER: You say we are staying until what,
10:00?
DR. POWERS: There is nothing sacred about 10:00.
MR. KARWOSKI: In the interest of time, I will try
to finish these last few slides. The last thing I wanted to
say about POD is there is a lot of issues. Some of those
curves showed it. We don't have 100 percent probability of
detection. It does, you know, 5 percent degradation for
some of those mechanisms.
Some of these concerns have already been raised.
Lab data versus field data. Is the field data -- or is the
lab data representative of the signals in the field? If
not, is it appropriate to use it.
DR. CATTON: When you say lab data, is that a lab
created defect as well, or is it just a lab measurement of a
pulled tube?
MR. KARWOSKI: It is like a model boiler specimen,
a laboratory created defect, like a stress corrosion crack
developed in a model boiler. Is the signal that I get from
that representative of the signal that I would have observed
in the field at that location?
DR. POWERS: Ken, you didn't discuss -- one of the
documents that we got, and I cannot remember which one,
spoke of a rather elaborate set of measurements on a steam
generator that had been removed from service at Surry, and
in which there was quite a lot of effort to address the
issues of probability of detection and whatnot. You didn't
bring that data up. Is there any particular reason why?
MR. KARWOSKI: The only thing, the Surry steam
generator, and Dr. Muscara can correct me if I am wrong, the
dominant degradation mechanism there is wastage and pitting,
and as a result, a lot of that was probability of detection
curves. Although they may be appropriate, it is not the
degradation mechanism of interest.
As part of the program, they did do that stress --
laboratory grown stress corrosion cracking mini-roundrobin,
and that is where the .6 came from. With that said, towards
the end of this presentation, I am going to have Dr. Muscara
present the results of another roundrobin that is going on
now. Results should be available at the end of this year,
sometime next year, and he will describe that program, which
is probably more relevant to inspection capabilities today.
The other thing is -- another POD issue is the
overall system versus technique capability. When you
analyze some of these tubes, it is part of the qualification
program. Assume they are all field data that you are
getting. These tubes are analyzed by a lot of people under
controlled conditions and get a much more detailed
evaluation than necessarily a production inspection where
the analyst is asked to look at thousands of tubes, you
know, in an outage.
So the question is, what are these curves really
measuring? Are they measuring what the technique is capable
of doing, or is it measuring the overall system, how the
analyst is going to perform under actual field conditions?
The other POD issue is, how pertinent is this
probability of detection curve to the plant that I am
applying it to? There are plant-specific circumstances that
may affect detection. Do I have copper deposits? Are my
deposits uniform such that I may be able to mix them out?
Are they spotty, causing problems in detection? Do I have a
lot of denting, a lot of noise from a variety of sources?
That is another issue with respect to POD.
The other is, how do I evaluate this data, and
what curves do I fit through it? Is it a function of
average depth, maximum depth? Do I use the log logistic or
what type of curve do I use to fit the data?
The other issue that I have put up is false calls.
When you do these inspections, when you pull a tube and you
look at it, you may tend to over call with respect to some
of these indications and say there is something there, but
then it is not there. Does that count against you in the
probability of detection calculation? Well, maybe it should
if you are not using the exact same criteria in both
applications. Those are just some of the issues that we
have with respect to the POD curves.
The first part of the presentation, detection.
Can techniques detect degradation? The next part is sizing,
and I am going to go through this relatively quick and just
provide some examples. There is a lot of examples in your
handout.
The first thing I would like to point out is a
technique may be qualified for detecting a form of
degradation but may not be qualified for sizing it. And
that is why I have the bullet -- Utilities routinely use the
bobbin coil for detection and other probes such as like a
rotating plus point probe for characterizing it. And sizing
can be in terms of length, depth, voltage. It depends on
what type of correlation you have for your structural
leakage integrity.
What is considered qualified with respect to
sizing? In the past the industry has used the root mean
square error approach. There is limitations to that.
Probably a more appealing approach might be just
understanding the uncertainties in your technique and then
applying those uncertainties in your condition monitoring
and operational assessment.
With respect to what is considered qualified,
generally, the bobbin coil is considered qualified for
sizing wall loss type of indications. The bobbin coil is,
quote-unquote, qualified for sizing via voltage for the
Generic Letter 95-05 methodology. And the NRC has approved
the use of the plus point coil for depth sizing degradation
at dented tube support plates, primarily axial PWSCC.
There is several pages in your handout with
respect to sizing curves, how well techniques do with
sizing. Once again, this is industry submitted information.
There are some issues with respect to this. I will just put
a couple of them up. If you have any questions, I will be
glad to answer them.
Here is a plot of the estimated throughwall depth
versus the true -- or throughwall depth as a result of
metallography. This is for wear indications and, in
general, you have a pretty good -- you can estimate the
depth of these indications pretty well, and that is
generally accepted throughout the industry.
And there is various other correlations. I don't
want to spend too much time on them given that we are
already behind schedule. But this is for sizing sludge pile
ODSCC, destructive exam maximum depth versus NDE depth, and
you can see that there is a lot of scatter in the data. And
there is various -- in your handout, there is various
correlations with respect to how well different techniques
can size specific degradation mechanisms in terms of length
and depth.
The burst characteristics of a tube depend,
basically, not just on, you know, a single parameter of
length or depth, but more a function of a composite of the
crack and there may be a limiting form -- a limiting portion
of that crack.
This plot here is a plot of a circumferential
crack located at an expansion transition which was pulled.
The metallography, this tube was pulled in the '95-'96 time.
The metallography is shown here. This was a deep
circumferential crack with some ligaments in between, two
deep portions of the crack, I guess it is wrapping around,
but there is a ligament there. That is the destructive exam
results. If you look at what the NDE analysts called this,
they significantly undersized portions of this crack.
DR. POWERS: Maybe clarify what the vertical axis
is.
MR. KARWOSKI: The percent through the wall.
DR. POWERS: Throughwall, yeah.
MR. KARWOSKI: So this is the crack profile.
Basically, it is 100 percent through the wall, in this
region there was a ligament, it went back to nearly 100
percent and then there was -- and this using an 080 pancake
coil, and that is just sizing of indications,
circumferential ODSCC at the expansion transition in '95-'96
timeframe.
I just point that out to show --
DR. CATTON: This doesn't say anything about the
cross-sectional area.
MR. KARWOSKI: Cross-sectional area.
DR. CATTON: Well, I mean what is the flow area
through a crack like this?
MR. KARWOSKI: No, it doesn't say anything.
Typically, the eddy current data --
DR. CATTON: You just take a slice?
MR. KARWOSKI: Well, they look at the various
ligaments and they develop, you know, somewhat of a
composite type of crack. But, you are right, it is a slice.
There is a couple of more plots of that same tube
with different types of probes. But just to show you a
little difference, one of the qualified techniques, or a
technique where we approved an alternate repair criteria,
here is a tube that was sized with a plus point coil for
primary water stress corrosion cracking at dents. That is
not listed on here, but that is the type of degradation.
And, in general, you see the destructive
examinations and the solid symbols, and you have the
analysts. Much better agreement with respect to both the
length and the depth of the degradation. And there are
several examples of this type of degradation in the handout.
DR. POWERS: And you said this was primary water
stress corrosion cracking?
MR. KARWOSKI: Primary water stress corrosion
cracking at dented tube support plate elevations.
There are many factors that affect the ability to
size, and a lot of them are similar to those with respect to
detection. What technique, what type of coil, the location
of the degradation, frequency. What are the interfering
signals? Noise.
Plant-specific considerations play a role.
Typically, the qualification, though, is done generically.
Licensees, however, need to assess whether or not that
technique is applicable, whether or not they can use that
generic qualification at their plants.
One of the things that some of the tube pulls
programs that have been done revealed is that, although you
can't necessarily determine the quantitative size of the
degradation, the general severity can be inferred from the
eddy current results. And what this allows licensees to do
is, although they -- was that me?
DR. POWERS: Your battery is probably going out.
MR. KARWOSKI: What this allows licensees to do is
to select some of the more limiting tubes for in situ
pressure testing to determine whether or not the tubes meet
the required structural and leakage integrity requirements.
There are a number of questions with respect to
the factors affecting the ability to size as a function of
different parameters. Crack orientation. Basically,
circumferential cracks are located in locations where there
is geometry changes that typically will make it more
difficult to size. Isolated cracks versus crack in
clusters. The problem there is the eddy current coil has
such a width that it may not be able to discern some of the
individual cracks in a cluster of cracks.
With respect to plugged cracks or cracks occluded
with crud, once again, it depends on the nature of the
deposits and crack location relative to support plates. The
more interfering signals that you have, the more difficult
it is to size.
DR. POWERS: Well, you have indicated a couple of
times that the plugging material affects it depending on its
particular nature, whether it is ferromagnetic conductive
or, I guess, if it is insulating sludge, it is just like air
getting in there. What do you typically have? Or is there
a typical?
MR. KARWOSKI: With respect to inside the crack,
it tends to be very tight and there is not really a lot of
deposits inside the crack. Most of the deposits are around
the outside of the tube. Okay. Those deposits, if they are
uniform, you know, if they are non-conducting, they may
contribute noise which will make it more difficult to
detect.
If they are conducting and they are uniform, you
probably have a better capability to detect the flaw that if
the deposits are conducting and they are not uniform,
because then you are going to get a lot of individual
signals where you won't be able to discern what is noise
from what is a defect.
Does that answer the question?
DR. POWERS: Well, I guess the one I am interested
in is the cracks are tight, there is very little material
within the crack, but surely there must be some material in
there. I am sure there is oxide coating on things.
MR. KARWOSKI: Right.
DR. POWERS: If indeed these are oxides or
spinels, they can't be -- at least any ferromagnetic.
MR. KARWOSKI: Right. And so these deposits will
interfere, but, in general, I have shown you some of the
data that the industry have presented. In general, that
will come out in whatever sizing curves that they develop.
So it will affect it, the magnitude.
DR. POWERS: Well, I guess what you are telling me
is, whereas, in principle, the cracks can be affected by the
crud that is in them, or around them, in fact, it is either
accounted for in the empirical determinations or it is just
not a very big affect.
MR. KARWOSKI: That's correct, that would be my
interpretation.
I have presented basically industry data with
respect to detection and sizing. The Office of Research has
a program at Argonne National Lab which is also looking at
the ability to detect and size flaws. So at this point I
would like to turn it over to Dr. Muscara, who is going to
discuss some of the work being done at Argonne.
DR. POWERS: I have some experience with the work
Argonne and we won't hold that against you.
DR. BONACA: Just I have a question I would like
to ask you before. Regarding the information distributed
this morning.
MR. KARWOSKI: Yes.
DR. BONACA: For those correlations. Is it all
field data or is it also laboratory data?
MR. KARWOSKI: It consists of both field and
laboratory data.
DR. BONACA: I looked for information on
separation of the two, I couldn't -- maybe I have to look
deeper.
MR. KARWOSKI: Some of the plots may discern
laboratory from field. I don't know if those do. But if
not, we can provide you the raw data that shows you which
ones are model boiler specimens and which are field.
DR. BONACA: A little understanding of --
MR. KARWOSKI: The relative counter.
DR. BONACA: -- how preponderant one group or the
other is over the other.
MR. KARWOSKI: Right. There is a lot more field
data now than there was in '95, but we can provide that to
you.
DR. BONACA: Okay.
DR. CATTON: Joe, what is a model boiler?
MR. KARWOSKI: When I said a model boiler, it is
just -- how should I? It is a means of creating laboratory
cracks, where you basically subject a tube to accelerated
corrosion tests, or accelerated corrosion, so that you can
develop a flaw in a subsequent test.
DR. CATTON: Boiler, okay. And what are added
tubes?
MR. KARWOSKI: Added tubes. Oh, with respect to
those correlations.
DR. CATTON: Yes.
MR. KARWOSKI: The industry updates the database
from year to year, so the added on that means we have added
new data from the previous one. So what they are doing is
assessing the changes with respect to adding the new data,
that is what those curves are referring to.
DR. CATTON: Okay.
DR. MUSCARA: Thank you, Ken.
I am not sure whether I should stick with my
viewgraphs. I mean I have heard a lot of things that I
would like to respond to, but I know we probably won't get
into that. But before I do get to the viewgraphs, there are
a couple of points.
I was really shocked yesterday to hear that in
NUREG/CR-2336, we had data on the probability of detection
that the tube support plate location, I think it was set at
the tube intersection, but tube support plate location. I
was surprised mainly because I had planned that work and
managed it for many years, and we did no such thing. There
was no work done that evaluated POD at the support plate.
And Dr. Powers was correct, we did a lot of work
on the Surry program to quantify the capability of inservice
inspection for the Surry generator. In general, we had, as
was mentioned earlier, mostly wastage and pitting, and we
did a thorough job of evaluating that type of degradation.
We had a number of industry teams inspecting the
generators, the same way that they inspected field
generators, and we got some valid data.
Well, this work was done essentially -- well, the
entire Surry work was done in the time period between 1982
and 1986. We did have lots of cracking in the Surry
generator at the support plate location due to the gross
denting that was going in this generator. But at that time
inspectors refused to give us information about support
plate, they knew they could not inspect in that condition.
The denting we are talking about in those days was
much, much more gross than what we are seeing today. So we
had no attempt at evaluating probability of detection of
cracks to support plate.
MR. BALLINGER: I have got four PNNL reports, one
of which does contain some stuff on POD, on a little test, a
roundrobin looking test where they did it.
DR. MUSCARA: Yes. Yes.
MR. BALLINGER: And that was -- it was in a report
related to the Surry generator.
DR. MUSCARA: Yes. Yes. It was done during that
time period. This is what I was getting to. But as far as
evaluating POD at the support plate, that wasn't done.
Because what did we do? In that time area, we were, of
course, beginning to experience cracks, stress corrosion
cracks, let's say the modern day stress corrosion cracks,
where we have families of small cracks with ligaments, not
the planar stress corrosion cracks that we experienced early
on in the life of these generators.
So we decided that we needed to do something to
provide some data on stress corrosion cracking besides
having the very thorough work done on the wastage and
pitting. And so we did run a small roundrobin. Now, this
roundrobin contained 17 samples. There was one sample at 85
percent throughwall stress corrosion crack, one sample was
throughwall. Twelve of the 17 samples had cracks above 40
percent, but many were between 40 and 50, the rest were
below 40 percent. So it was not a great deal of work.
We did include in the sample set, samples that had
copper-coating, because we know that this complicates the
inspection. And in four of the 17 we had support plate
simulation with the specimen. Again, the intent wasn't to
really evaluate POD at the support plate location, the idea
was to get an idea of how teams at that time might be
performing on stress corrosion cracks.
So the numbers that were given for POD at the
support plate location, yes, they were somewhere between .27
and .5. A little bit about what kinds of teams looked at
this, at this information. We clearly had sent a small
mockup to a number of inspection agencies and they used
their field teams. We also asked them to look with their
most advanced techniques and their researchers, not
necessarily just the fuel technicians.
We had looked at the bobbin coil on these samples.
Pancake probes were beginning to get developed in those days
and beginning to be getting used, so we also had pancake
coil inspections. And there was one probe that was under
development which never really got into the commercial
arena, and this was called, in our report, we called it an
alternate bobbin call. This was what was called a segmented
coil. The idea here was to segment the coil so that you
might get some information about the location of the flaw.
As you know, the bobbin coil, the encircling coil
essential integrates around the circumference of the tube,
so it gives you some information about the flaw, but not
necessarily the location of the flaw.
As a matter of fact, we do have some probes here
that we could send around. Thank you, Bill.
This is a typical bobbin coil, you see the
encircling coil there. And this probe head has three coils
on it, there is a plus point and two pancake coils. The two
pancake coils are different sizes. The larger size gives
you more sensitivities, a larger signal, but less
resolution, so they have the combination of these two,
depending on whether you are interested in resolution or
sensitivity.
I guess I might mention, we talked about the
bobbin coil, and, you know, everything is bad about the
bobbin coil, it doesn't detect circumferential cracks.
Well, there are some things that are good about the bobbin
coil. Of course, one was already mentioned, it is the
speed. But if you are looking at a small level signal,
let's say, from an actual stress corrosion crack, the bobbin
coil actually gives you a larger signal than these pancake
probes, which these are small coils.
They don't have the resolution. Of course, if you
are trying to evaluate the size of the flaw, if you are
trying to map the shape and size of the flaw, you are better
off with these pancake probes. But from a point of view of
detection, especially for small signals, the bobbin probe is
a better probe. So in a voltage based criterion, where we
are looking at accepting flaws that are less than two volts,
it makes sense to use a bobbin coil. You know, we do get
more sensitivity to those kinds of flaws.
Again, before I get to the viewgraphs, there are
several things we could say about POD and how people do
these POD tests. Just a couple of points in general, we
have had the experience of doing roundrobin inspections for
many different components, not just steam generator tubes,
but also piping invessel. So we have done a lot of work on
evaluating the POD of different techniques, say,
ultrasonics, eddy current for the different inspections.
And we also have been involved in the work not only in the
States, but international work.
And when you talk about the reliability of
inspection of POD, it is made up of at least two components.
We like to talk about reliability of inspection in terms of
the NDE system capability, and the system is really made up
of the equipment, the procedure and the personnel, and each
taken together give us some idea of the reliability of the
inspection process.
So a lot of the work that we do in evaluating POD,
of course, is laboratory work, so, in a sense, you know,
people know they are under test conditions, so you get more
or less an upper bound of what you might expect for a field
inspection. Very often we do not have enough samples in
sets of samples. If you are trying to evaluate POD, a high
POD at a high confidence level, you need hundreds of flaws,
not what we do in qualification where we are looking at a
handful of flaws.
We often use notches, which, again, are not
realistic when one is trying to do POD data.
In our work that we have done, where the work has
been robust, we are really trying to determine reliability
of inspection, we find that you never get to 100 percent POD
when you use the system, the person, the procedure, the
equipment.
Now, we can break this down into capability.
Depending on the physics, the equipment can have the
capability to detect 100 of the flaws it is supposed to
detect. But then we put this into the hands of an inspector
and the human reliability comes into play, and we do not get
100 percent POD. Regardless of what some of the data shows
us when you look at three, four, five, ten flaws, detect all
of them, therefore POD is 100 percent. That just is not the
case when you use hundreds of flaws and inspection teams
outside of the laboratory, more on a situation like a Surry
roundrobin.
I have seen information on POD, for example, where
information is shown as POD going up to 100 percent at about
the 50 percent level of degradation. But when you question
how this information was obtained, the number of specimens
is somewhere around 30 or 40. The mean size in the sample
set was 27 percent throughout wall depth, the maximum depth
was around 36. Yet they show us a curve of POD, you know,
at 50 percent throughout flaw, POD is 100 percent and it
stays 100 percent from thereon out.
Well, in questioning about how this was developed,
we used the logistic curve fit. So, you know, the data set
is down -- 27 percent is the mean depth of the flaw, one
flaw at 36 percent, and with logistic fit, we wind up having
100 percent POD.
DR. POWERS: How do you use a POD when you are
talking about a production process? Suppose I have got a
run through a particular tube and that tube has -- in some
way we know absolutely that the tube has five indications in
it that are, let's say, 50 percent throughwall, okay. Now,
and I have a probability of detection at 50 percent
throughwall of, say, 80 percent with a 90 percent confidence
level. And I ask what is the probability that my analyst
will find all five of them?
Is it 80 percent times 80 percent -- 80 percent to
the fifth power or something like that, or is it another
number?
DR. MUSCARA: I guess, like Ken said before, I
have to defer this to our statistician. But we have done
the statistics and determined the number of flaws that you
need to evaluate POD at different levels. And the way we
have set up some our roundrobins are based on this. When we
report our POD data, they have a confidence level that is
statistically based.
I Ken was showing some of his viewgraphs, 13 out
of 13 giving 100 percent POD, but then we apply the number
of samples that were use. The 90 percent confidence on that
was really 80 percent.
DR. POWERS: What I am driving at is -- I think it
is just what you have been saying. If you give me a small
set of samples to do, and a relaxed period of time to do it
in, then essentially I am doing each indication alone, it is
a separate experiment and I get a particularly probability.
But now when I am running this detector through
the tube in one big operation, now the question is not
individual cracks but what is the probability I will detect
all five of the indications that are known to be there? And
is each one of them an independent event, or do I get a set?
DR. MUSCARA: Yeah. In fact, in order for the
statistics to work, it has to be an independent event. And
when we evaluate our roundrobin data, we essentially divide
up the test section into what we call grading sample,
grading units. And there are certain requirements for the
flaws. For example, in order for the numbers to be
independent, the signal from one flaw cannot be interfering
with the signal from the other flaw. So there are a number
of rules that are set up to separate these so that when we
do run the statistics, we get the correct answer, that each
measurement is essentially independent.
DR. POWERS: But are they independent when I am in
a production run? I guess that is the question.
MR. STROSNIDER: This is Jack Strosnider. Just if
I could interject just for a second, Joe, I guess. The way
this actually happens in the field is typically there is two
reviewers, and they are working independently. And then
depending upon whether they agree on their calls, it goes to
a third reviewer for disposition.
And we have had a number of discussions with the
industry, and I am not sure how consistently this is being
done now, but one of our concerns, first of all, you have
the question, you have got two analysts looking at the same
signal. Can you that their probability of detecting the
flaw is truly independent? Now, you can make that
assumption, but, in fact, there may be noise or something in
the signal that makes it difficult for both of them. There
is probably some dependency there, but hard to quantify.
The other issue that has come up in some of the
reviews we have done is after they go through their
comparison, they give it to this third fellow who is usually
a more senior level analyst, and he makes the decision. So
it comes down to, you know, what he is deciding.
So, I don't know, and staff can fill me in, that
there has been any real methodical study where we can tell
you what the probability of missing an indication is even
with a qualified method. There is some finite probability.
It is, you know, it is not a foolproof process, that is for
sure.
And without going into a lot of detail, I would
just say, when you recognize that, you have to recognize
that inspection is just one layer in defense-in-depth that
is applied to managing steam generator tube integrity. You
have leakage rate limits. You have the fact that it is a
design basis accident. You have your operating procedures,
and so on.
So, we recognize that inspection is not, you know,
not 100 percent reliable.
DR. POWERS: Well, I guess I am just worried about
the theoretical issue of whether, in a production run,
looking at just one tube, it is a set of independent
indications or it is a collection. And how you do the --
how I divide the probability of detection, I think the way
you are applying it is all independent isolated events.
MR. STROSNIDER: Yeah. Well, let me give you one
other thing there just to add some perspective on this. It
might require some more discussion later. But we talked
about the voltage based criteria, and you heard some of the
discussion about how the probability of detection, et
cetera, is applied there. Ken mentioned one other plant
that has an alternate repair criteria for primary water
stress corrosion cracking at the tube support plates. And
they -- and I guess maybe I am jumping a little bit ahead,
because it is not just POD, but it is sizing. And those are
the two alternate repair criteria, we pretty much rely on
that.
Most of the operational assessments that are done,
all right, there is not a whole lot of calculating and the
kind of stuff that we have been talking about, it is
basically, do the condition monitoring at the end of the
cycle, which relies to a large extent on the in situ
testing. And if you show that you meet the margins when you
do that, the assumption pretty much is that your probability
of detection and growth rate is such that things remaining
the same, without any, you know, significant changes, you
ought to be able to operate the same length of time again.
Like I said, there is always the possibility that
new things show up in between. But I think people have the
perception that, you know, that every review and every
operational assessment that is done, that people are going
in and using all these numbers and stuff, and that is not
really the case.
But it does come up, it came up in the Indian
Point review. It came up in the Arkansas review, it came up
in Farley. We will talk a little bit about those tomorrow.
So I don't know if that is helpful, but that maybe
provides a little perspective on how it is actually applied.
DR. POWERS: Well, certainly, it provided a
perspective on the overall problem, the challenge that Joe
has.
MR. STROSNIDER: And just to follow up on that, as
we move more toward these risk-informed and start doing more
risk-informed amendments, there will be more reliable on
this. You know, part of what we are trying to get across,
working with the industry to get, is reliable data that can
be used in those type of analyses. And there is certainly
room for improvement at this point.
DR. BONACA: I would like to ask one more question
on this issue. When they ran the bobbin coil and they get
these five signals in a tube, first, that will have to the
one of deciding what kind of indications these may be. For
example, they are not all going to be one type of defect.
There is going to be different types of defects. So I
imagine that they have some techniques by which they control
these defects in different bins.
And I imagine that, for example, correlating one
defect with the position of the plates, and so on and so
forth, will help the selection, but, you know, it is not
clear to me how this complicates the process. I think we
had an overhead that showed that there was some
consideration of the process as one complicating factor.
But I imagine it is a complicating factor.
MR. STROSNIDER: I guess you are saying you are
trying to understand exactly what the type of degradation is
that you are dealing with.
DR. BONACA: Yeah. Because I mean all we have
talked about in these past two days is one type of defect,
and characterizing it, and we have seen a distribution of
that versus voltage. But, really, this is the process by
which they are identifying all the defects.
MR. STROSNIDER: You can get some information from
the eddy current, but, as we said, you know, it is not -- it
has limitations in its ability to characterize a defect.
You also, you know, based on operating experience, and as
you suggest, the location of the defect, draw some
inferences from that.
When people find some new things, or some things
that are unusual, two pulls are the way to get some solid
information, but there is a challenge here.
One other thing I would add, too, which may not
have come across in all the discussions we have had so far,
is that, aside from these alternate repair criteria that we
talked about, which are relatively few, the industry
practice it to plug on detection. So when they find a
defect, a stress corrosion crack in particular, they are
going to plug that.
Now, thinning and that sort of thing where they
have some qualified and reliable sizing method, that is not
true, but for IGSCC, basically, plug on detection.
DR. MUSCARA: There are codes that the industry
uses for characterizing flaws, but, generally, they are
based on past experience. A lot of it is based on location.
But, generally, you can discriminate between large volume
defects and small volume defects. We can discriminate
between cracks and wastage or pitting. And then within the
crack regime, we can discriminate the circs from the axials.
So there is some capability even with the
detection probes. And then, of course, if you are doing
more careful work with some of these rotating probes, you
have additional capability for characterizing the flaws.
DR. BONACA: Yeah. I just was wondering, for
example, if the fact that you are going through and picking
up a lot of signals could confuse this detection ability.
DR. MUSCARA: Right. With the detection, the
problem is if the signals are close together, then it could
confuse it and you could mess up the statistics. In the
tests, of course, we do make sure that we have independent
measurements.
In the field, if you are finding different kinds
of flaws very close by, that confuses the issue. But,
generally, I don't think that is the case. We find a
cluster of flaws at the support plate, we know the type of
flaw they are, they are in a certain zone. The next flaw
you would find maybe at the next support plate. So they are
not really -- one is not really affecting the other.
I guess just to finish up what I had started on
the NUREG-2336, the loan numbers were developed when we were
using this, you know, I mentioned there was an experimental
probe, a segmented bobbin coil, so that is where the 27
percent came from. That has never been used in the field.
I guess also I should mention that in that test,
we were using single frequency eddy current at the time. So
when you look at the data from the normal bobbin coil and
the pancake coils, the average POD was, as was stated
earlier, was .63, I think we are using .6. I think the
maximum we found in that small roundrobin was about .75.
But one of our main objectives, let me just
mention it, the Surry work was I think very useful, we had
very valid data. That was part of an international program.
The Surry part of the work cost $17 million, and the flaw
types of a different nature these days, we need to do
similar kind of work, but we can't afford to spend another
$17 million to get these POD curves.
So what we are trying to do is set up a mockup and
an inspection process that mimics what goes on in the field.
And so we are trying to set up this mockup so that it has
the kinds of conditions one runs into in the field, so that
the flaws are typical of what is in the field, and so that
the inspection process is also conducted according to the
qualified procedures and so on. And I will get into some of
this work.
Unfortunately, I will not be presenting a lot of
new results right now. We will have information by the end
of this calendar year. We are in the midst of conducting
the roundrobin. We are trying to keep this a blind test.
There are some other teams that we need to bring on board,
so we cannot release a lot of the information, but I can
give you some trends.
We also mentioned earlier some advanced
techniques. I wasn't planning on talking about the work we
are doing on advanced techniques, but we are doing research
both in characterizing the reliability of current inspection
methods, most of this is with the mockup and the roundrobin
testing, but we are also doing work on advanced eddy current
techniques, in particular, data analysis procedures bring
some of this in, and that we are also using this for
characterizing the mockup.
And I have probably mentioned already what is in
the first viewgraph. The purpose, again, is to evaluate the
reliability of current day inspection, both with respect to
probability of detection and sizing accuracy, and we will be
using a mockup.
I guess while I am going through the viewgraphs, I
also mentioned a couple of items. When we talk about
qualified techniques, you know, that is very soothing. When
we talk about something being qualified, we think it must be
good. I think we need to pay particular attention to what
we mean about qualified techniques with respect to what is
being qualified for inservice inspections.
When we talk about qualified detection techniques,
the technique can be qualified if it passes a particular
test. There are a certain number of samples that are
involved. Normally, there are not a great deal, a number of
samples. But the passing criteria is that you need to get
80 percent of the flaws at 90 percent confidence level for
flaws that are 60 percent deep and deeper. So if we are
talking about a 40 percent plugging criterion, do I always
know what the probability of detecting a 40 percent flaw is
when the qualification is at 60?
In the sizing arena, when we started doing
qualification on sizing, the criteria had been 25 percent
root mean square error. We are no longer using that. So
what is implied in a qualified sizing technique is that the
process has gone through the system, a test has been
conducted, but there is no passing criteria. But they do
record how the person and the system performed. So a
qualified technique would be something that gives you a
sizing accuracy of plus or minus 50 percent. If it has gone
through the system, it is qualified.
So we need to understand that the qualified
doesn't necessarily mean it is very good, but at least we
know how it performs. And then that information, of course,
is used in the operational assessments, and that is the
important point. But we shouldn't be left with the idea
that a qualified sizing technique may be a very accurate
technique.
Just a very brief description of the mockup, we
have essentially 400 tube openings. Each tube is made up of
nine test sections, so there are nine individual one foot
sections. They may have a flaw, they may not have a flaw.
But there is the option of having 3,600 test sections in
this mockup. At the top of the mockup there is a three foot
run out section, and that is there so that -- well, the
probe doesn't fall out of the tubes when we do inspection,
but more importantly than that, we want to make sure that
when the probe hits the first sample, that the probe is up
to speed, so it has a constant speed throughout the
inspection.
I mentioned they were trying to make this mockup
realistic. We have literally hundreds of flaws. Again,
since it is a blind test, I don't want to mention how many
hundreds, but it is several hundreds of flaws. The types of
flaws we have are mostly stress corrosion cracks from the
ID, axial, circumferential. We have some IGA. There are a
few EDM notches and a few fatigue cracks.
We also try to reproduce conditions in this
generator, so that besides the straight sections of tubes,
we have tubes that are rolled into tube sheets. We have the
same roll transition in these tubes as we have in operating
plants. There are dents in the tubes. We have sludge
piles, we have magnetite. So that we have tried to
reproduce the conditions that are important that affect an
inservice inspection signal.
MR. BALLINGER: No U-bends?
DR. MUSCARA: No U-bends, correct.
DR. CATTON: 22.2 millimeter diameter is 7/8ths of
an inch?
DR. MUSCARA: That is three-quarters, right next
to it. So, yes, the other items that we do have an actual
carbon steel support plate, we have three simulations of
these in this mockup. So that is the same size as a support
plate out in the field.
As I mentioned, we are trying to mimic the
inspection process that goes on out in the field, and we
know a lot about that inspection process. Our researchers
at Argonne know something about that. But we really wanted
to make sure we were doing this right, so we put together an
NDE task group. And the idea here was that we wanted to
have some input that actually do these inspections, people
that develop the inspection plans.
So we put together this task group and the members
were from Argonne, from NRC, from EPRI, from FDI, which was
the old Babcock & Wilcox, ABBCE, from Zetech, which is a
major inspection company, provides inspection services and
also equipment, Westinghouse, Northern States Power,
Commonwealth Edison and Duke Power. We met a number of
times to discuss this test. But the main input we had from
the members was that we wanted to know if the signals that
we had from the cracks in this mockup are typical of what
they see out in the field, because we wanted to make sure
that the cracks are prototypic both from the point of view
of the morphology, from a metallurgical point of view, and
also from the eddy current signal point of view.
And so we have compared these cracks to signals
that you get out in the field, and they are typical. We
also compared them to the metallography of stress corrosion
cracks that we get from pulled tubes and they are quite
typical. Of course, stress corrosion cracks, if you have
seen one, you have seen they are all, but there are minor
variations, and we cover the range of the stress corrosion
cracks that you do notice in the field.
Now, we do have in the mockup cracks that come in
clusters. We have single cracks, but many of them are
today's type of crack where we have small cracks with
ligaments in between, and these are distributed around the
circumference in one more than crack and also axially along
the tube.
DR. POWERS: How did you make these cracks?
DR. MUSCARA: Well, these cracks were made in the
laboratory. When we started out, we were using autoclaves,
high temperature caustic solution. That got to be too
time-consuming and too expensive, so we have been working
for a number of years, one or two years, to just develop
methods for coming up with these cracks. We essentially
heat treat the tube so it is sensitive to cracking and we
conduct the cracking at room temperature in one more
solution of sodium tetrathionate.
DR. POWERS: Tetrathionate.
DR. MUSCARA: And we can get cracking with this in
time periods of the order of a day to three or four days.
Now, all these cracks are very well characterized.
I mean we know where the flaws are because we are
introducing them. But they undergo a battery of tests where
we used advanced NDE techniques, whether it is ultrasonics,
mostly eddy current, die penetrant. We do a lot of work to
characterize these cracks before we accept them for the
mockup.
Again, we want to make sure that they are
realistic and sometimes we aim at certain kinds of cracks.
We can produce fairly closely what we need, but it is a
random process. So, you know, sometimes we reject some of
the cracks, they may be too wide open for us.
Well, in addition to assuring that the cracks are
typical and the conditions are typical, we also wanted to
make sure that the roundrobin is conducted in a manner
similar to inspection conducted in the field. So we
effectively treated the generator -- in the field, of
course, there is an owner of the generator, and the owner is
responsible for what goes on with this generator. The owner
is responsible for coming up with the inspection program.
So we assigned Argonne National Laboratory the ownership of
the generator, so they act as the owner, and they are
responsible for developing the inspection program.
Now, when developing the inspection program, a
number of things are taken into account. For example, the
owner is responsible for doing a defect analysis, and so
they are required to sit down, determine for their plant the
kinds of degradation they have experienced in the past, and
determine for sister plants what kind of degradation they
are experiencing. And then they are required to make sure
that the techniques used for inspection match the kinds of
degradation that they are experiencing.
So, they are supposed to be using qualified
people, but, in addition, the personnel needs to be
qualified at the plant site. So they need to take a
plant-specific examination, both a written examination and
an examination, an actual examination of inspecting data
from their plant from the past where they know what the
situation of the flaws are.
So a lot of information was gathered, a lot of
documentation was written, very similar to what is conducted
in the field. And our task group was very helpful in
providing us with a lot of this information. So we had a
lot of information, for example, on inspection plans that
are used at actual plants, and so we mimicked our process
along that information for the mockup.
For an actual inspection, there is usually at the
utility is a Level 3 inspector who is responsible for
approving the inspection program. And, similarly, for this
mockup, for our roundrobin, we had the task group had the
responsibility to review all the information and approve it,
that the techniques we are using match the requirements and
that they are appropriate for the kinds of degradation that
we have in the mockup.
So, what I really wanted to stress is that when
you do see our information, come December or January, that
you do have the feeling about how the work was done. This
is not just a laboratory test. You know, clearly, I can
mention laboratory tests and roundrobin that we have done in
the past, some international work where they use all
notches, 100 specimens, 95 which were notches, five were
cracks, and they tried to develop UOD curves from that.
This work is laboratory, but it is a mockup and it
is trying to reproduce the conditions of the field, and
inspections are conducted in a manner similar to what is
being conducted in the field.
After assembling the mockup and doing a lot of
work on our own to characterize the nature of these flaws,
we started the actual roundrobin in February of this year.
We are dealing here really with an analysis roundrobin. In
the past, if you know the work from Surry, we have done both
what we call data acquisition and analysis roundrobin, and
we also conducted analyses roundrobin.
What we found is that the data acquisition,
regardless of who performs the data acquisition, you get the
same result. It is the same procedures used, the same
equipment, and the flaw is the same flaw, so that we found
very little variability in having many teams gathering the
data, the data was the same. So that we decided here that
we needed a run, an actually roundrobin, this is where the
variability in POD comes from, not from gathering of the
data.
So we had a qualified team from Zetech gather the
data. There was some oversight of the team, there was a
proctor present making sure that the data was gathered
according to the procedure. But then this data was given to
a number of independent commercial teams to do the analysis,
provide us the information about what they have detected,
and we also required information on sizing.
At this point I believe we have five teams that
have done the analysis roundrobin. We have incorporated
most of the major inspection agencies that conduct
inspections in this country. One clear exception at this
point is Westinghouse has not been able to participate yet.
I should mention that the program I am talking
about is an international program. The participants in this
program are Westinghouse, EPRI, Canada -- there is one more,
Korea.
So Westinghouse is very willing and has several
times scheduled to do the analysis roundrobin.
Unfortunately, they have been pulled away on other
activities. The last one they were scheduled to do the
analysis roundrobin for us in May and, unfortunately for us,
they got pulled away with Indian Point 2. They are now
scheduled I believe for November, and, hopefully, in
November they will do the roundrobin, and we can include
them with the set of information we already have.
Let me describe a little bit the teams. In
effect, as I mentioned before, we conduct it the same way as
the inspections are conducted in the field. So the
inspectors have been tested, they are qualified inspectors.
They have been qualified through the EPRI NDE Center. We
use a five person analysis team, and this is what is going
on today.
I also must mention that this technology is
evolving and it is improving fairly rapidly. Jack mentioned
we have two inspectors. In fact, when we planned the
roundrobin, there were three inspectors involved in the
team, now there are five. The description of the five
member team is that there are two analysts, we call them
primary and secondary, but, again, you know, they do the
same function. A secondary team doesn't mean it is less
qualified or the result is less important.
There are two independent teams that do the
analysis. In the past, if these two teams did not agree,
then a resolution analyst, who is a Level 3 rather than a
Level 2, will do the first -- the primary and secondary can
be Level 3. They are normally Level 2 and Level 3, but the
resolution analyst is usually a Level 3, and he decides what
the true call is if the primary and secondary can't agree.
That was the past.
Now, we are using five teams. So we have two
initial inspectors doing the analysis, primary and
secondary. There are two resolution analysts, and they have
to come to a consensus on the call. And there is a fifth
member of the team which is called the independent QDA or
the independent qualified data analyst. And this fifth
member is usually a member of the utility rather than the
inspection agency, not always, but usually.
So the makeup of our analysis team is made up of
these five inspectors. The true independent who looked at
the data, if there is something to be resolved, the
resolution analysts look at the data. And then finally, the
independent QDA has an opportunity to look at the data and
provide a final answer.
We have mentioned there are not too many sizing
techniques that are qualified, there may be one or two. But
we are requiring these teams to provide us with sizing
information, at least to give us the maximum size of the
degradation, and that is a sizing technique that is based on
the face angle of the indication. If you have to do sizing,
this is a typical method for sizing flaws. And it is very
similar, what we are requiring for the max size is very
similar to one of the qualified techniques for sizing.
I must mention also that, in addition, we plan on
getting a subset of this data to a number of commercial
teams to provide us with sizing information, not just the
maximum size, but we will ask them to do the entire mapping
of the flaw. This is not something that is required or is
qualified, but we are trying to get some idea about the
sizing accuracy also in this work.
In order to be able to grade or to evaluate a
roundrobin, we must know what the true state of the mockup
is, so we must know what the actual size, and type, and
location of the flaw is. Well, as far as location, we know
that. The most difficult part is to try and determine what
size these flaws are.
We want to be able to use this mockup in the
future for evaluating emerging technology. It is very
time-consuming and expensive to produce these tubes that
have realistic flaws. I guess, just to mention the mockup
itself, putting it together, making some of the flaws cost
over a million dollars. So these tubes are very valuable
and we do not want to destroy them all. So we are trying to
find some techniques for getting a true state of these flaws
so that we can then use in evaluating the performance of the
analysts.
And I will not spend a lot of time on this, but we
tried many techniques, including ultrasonics and high
frequency ultrasonics, land waves, all kinds of eddy current
techniques. Just to summarize, none of those worked that
well for all kinds of flaws.
We are concentrating now on one technique, which
is part of the research work we are doing not to evaluate
the reliability, but research that we are doing on advancing
data analysis techniques. And so right now we are
benchmarking a technique that was developed at Argonne
National Laboratory. It uses multi-frequency eddy current,
but in addition to the multi-frequency eddy current, we do
filtering, we do deconvolution. And we have developed a
rule based smart system, that is, we are incorporated into
this algorithm the kinds of things that the good inspectors
do when they do an analysis and try to decide whether it is
a flaw or not a flaw. So all of these rules have been
incorporated into this system. So one major aspect is the
multi-frequency correlations to flaw sizing, and that is
incorporated in this.
And what I wanted to show you next, just very
briefly, is some of the capabilities for sizing. We are
validating the technique by destroying -- by inspecting this
set of tubes, having the inspector provide us with the
mapping of the flaw, and then we are destructively
evaluating these samples to determine how well this
technique is working.
It is just an example of the result we get from
the technique. You have seen the kind of graph that you see
on the left before. A key aspect of this is that it has no
resolution. One of the key aspects of our rule based
automation calibration and deconvolution that we are doing
is to improve the signal to noise ratio. And the key
parameter in being able to detect and also size flaws is to
have a clean signal. So if we can reduce -- if we can
increase the signal to noise, that helps both the detection
and sizing. And one major aspect of this work is that we
are really reducing the signal to noise -- increasing the
signal to noise, reducing the noise a great deal.
So that shows the kind of information we get, and
on the right, we just show that you can section this
information, looking at the flaw either from a
circumferential, from an axial, or from a longitudinal view,
and you can get the profile of the flaw. And we are
evaluating how accurately we are doing this by destroying
samples.
Out of the 29 samples, we have finished that work.
All of those have been destroyed and compared to the eddy
current result. This is just a set of three samples that we
have looked at and compare the eddy current profile versus
the actual metallographic profile. I could spend some time
on how that is done, but if you have specific questions, we
can try and answer them, but in the interest of time, I will
move on. I just want to say we are doing a careful job of
metallographically evaluating the flaw.
DR. KRESS: How do you cut the tube?
DR. MUSCARA: Well, we don't cut the tube. The
best way we have found, that gives us very good information
that is more effective from a cost and time point of view,
is to pressurize the tube a small amount to open up the
flaw. We then look inside the flaw face, the tube is heat
tinted, so we know what the prior flaw is from the heat
tinting, and, also, it is an intergranular crack which is
different from any quote we might have had from the
pressure.
We take a digital picture, the picture is
digitized and we do digital analysis. And looking at the
light areas and the dark areas, we map out the flaw. It is
a very difficult and time-consuming process because the
flaws we have in some of these tubes are literally dozens to
hundreds of flaws, small flaws with ligaments. And we are
trying to evaluate those very carefully both with the
metallography and with the eddy current. And to my surprise
at least, we are doing a lot better than I thought we could
do with the eddy current technique. We are really getting
much resolution from this technique. And, as I say, we are
checking it out against samples, so know that it is real.
Well, this shows a comparison of three samples,
the eddy current by using this technique versus the actual
destructive examination. And you can see it is quite close.
In some cases, for example, the eddy current here does not
pick up the full length of the flaw, and that is fairly
typical. The probability of detecting shallow flaws is
small. So in these matters, if we turn to a length sizing,
we often undersize because we do not pick up the part of the
flaw that is shallow. That is a fact.
The interesting thing is that when you try to
evaluate these flaws and calculate a burst pressure, you
don't really need to know the shallow part of the flaw
because that does not contribute to the failure pressure of
these complex flaws. In general, what we found by running a
number of tests, and Bill will talk about some of this
tomorrow, but the portion of the flaw that is less than 70
percent throughwall usually does not participate in
determining the failure pressure.
So when we look at these flaws and we go through a
process of characterizing the flaw where we look at the
equivalent area or an equivalent flaw, length and depth,
because these are not rectangular flaws, and then use that
in our integrity correlations, we can estimate the burst
pressure very well by using these profiles, even though we
might miss the shallow part of the flaw from the NDE.
DR. KRESS: When you do the destructive test, is
there any chance that you change the flaw characteristics by
doing that? It looks like you --
DR. MUSCARA: Well, of course, there is always the
chance, but, again, we are careful. We do not open these up
a great deal, we just want to open them enough so that we
can look into the face of the flaw.
DR. KRESS: Yeah, and you can tell where you might
have changed it.
DR. MUSCARA: Right. And if we do change it, then
we know that that is different from the heat tinted area,
number one, from the intergranular nature of the crack.
DR. KRESS: And you can see what has changed.
DR. MUSCARA: So we can see. We do look at these
things on the scanning microscope, so that --
DR. KRESS: Oh, you look at them on a scanning
microscope.
DR. MUSCARA: Oh, yeah. Yeah, when we need to. I
mean some cases we don't need to. But in many cases we do
look at the surfaces on the scanning microscope before we
decide what the profile is.
Well, again, as I said, I wasn't going to -- I
really wanted to come in and show you some example PODs, and
we are not doing that, partially because we haven't fully
characterized the generator. We are still in the process of
evaluating the data. We are shaking down a statistical
package we have for conducting these analyses, a number of
reasons. And, also, you know, it is a blind test and I
really didn't want to have in the public some of these POD
curves.
But I can tell you, qualitatively, we remember the
Surry kind of information. We are not too far from that.
If you remember at Surry, we had some fairly high PODs for
large flaws, but never went to 100 percent. Real teams miss
flaws even though they are big. All they have to do is
blink while they are looking through the record. So that in
reality, some teams miss flaws. And we know in this case
also, some teams missed once in a while a large flaw. So
the POD does not go up to 100 percent, nor is it 60 percent
for the large flaws. So we are doing better than 60
percent.
I say in general I think we are doing quite well,
but you will see all this information in several months in a
much more quantitative way.
DR. KRESS: Is your objective to get a POD versus
flaw size?
DR. MUSCARA: POD versus flaw size as a function
of the technique that was used, which is qualified, so it is
POD as a function of the flaw type, the technique, the
location in the generator. We are going beyond that. I
mean in the past we looked at POD as a function of the
maximum depth. Well, maximum depth is not a really good
parameter for determining burst strength of these tubes when
we have a complex flaw. It is if it is a simple rectangular
flaw, but for these complex flaws it is not a good
parameter. So we will be looking at POD as a function of
other things.
One of the items that works very well, that we
find worked well for predicting burst pressure is this M sub
P. M sub P is a correlation factor, it is almost a stress
magnification factor that describes the stress at the
ligament of the flaw that is used for predicting burst
pressure of different types and sizes of flaws. So it takes
into account the geometry of the flaw. And Bill will cover
a lot of this development tomorrow.
But one of the important parameters here for
describing the severity of the flaw is M sub P. So one of
the things that makes a lot of sense to us is to try plot
POD as a function of M sub P. And it is not just the
research work, I mean even in the field, we are moving
towards, we are using this evaluation, we are using M sub P
for predicting burst pressures.
Besides the voltage criteria, and you have heard
there are a few other criteria out there, one of the most
recent ones is a criterion where you are actually using
length and depth of the flaw to predict its burst pressure.
And it is not just length and depth, in fact, it is what I
showed you before, it is the profile and how to calculate
the burst pressure of those tubes. Well, you need to have a
severity factor which is this M sub P.
So, you know, the laboratory work, yes, is leading
some of this, but it is winding up the field, and we are
doing those kinds of analysis. And, in fact, we have an
ultimate plugging criterion at the support plate and dented
region where we use this kind of an evaluation for
calculating a burst pressure and making sure that it meets
the 3 delta P.
And so with that, you come up with something other
than 40 percent, depending on the reliability of sizing, the
crack growth rate and the strength of these tubes.
So one of the things that makes a lot of sense for
us is to evaluate POD as a function of M sub P. And since
we are using voltage and we get that free, the voltage
always comes with the signal, we will be plotting, I am
sure, POD as a function of voltage for these different kinds
of cracks.
And then again, for the first time, we will have a
comprehensive data set where we know what POD as a function
of voltage is. When we looked at this POD of .6, it was as
a function of a handful of flaws of varying sizes. And
normally POD is a function of size, it is not POD is a
function of voltage. But we will have that information once
we are done with these analyses.
I am not sure if I should go through this. I mean
you can read it as well as I can. But we find is that the
POD for the larger flaws, or for the large segments of the
flaw can be fairly high, above 80 percent. Again, it is not
100 percent. We have missed sometimes large flaws, but it
is more than .6. And so there is -- we realize that .6 is
conservative, and in a voltage based criterion, .6 covers,
as we mentioned earlier, a number of things, not just the
POD but the crack's initiator in cycle, for example.
So the POD will have detailed data, can get fairly
high. On the other hand, it is very low for flaws that are
smaller than 40-50 percent, and that is not a surprise, I
think we expect this.
I think you can read the rest at your leisure.
DR. POWERS: Everything else is pretty much as
expected.
DR. MUSCARA: Right. Just very briefly, we have
been talking about sizing and the difficulty sizing. Sizing
is usually based on a calibration, so you have a set of
standards with different depths of holes or notches, and you
look at the face angle for each one of these notches, and
you have a calibration curve for sizing.
We also have indicated -- maybe we haven't, but
sizing ID flaws is more troublesome than sizing OD flaws,
and this graph shows you the reason why that is. If you are
looking at the ID flaws, that is the portion of the curve in
red. You can got from 100 percent -- from zero ID flaw size
to 100 percent through the wall size, and you are just using
up 30 degrees of face shift. So within 30 degrees, we have
the full span from nothing to throughwall. And when the
signals are complex and complicated by noise, and it is
difficult to pick out where one should measure the face
angle, then you get into a problem with getting good,
accurate sizing.
For OD cracks, they normally can be sized a little
bit more accurately. There is a larger span than covers
from 0 percent depth to 100 percent. But, at any rate, so
sizing normally is conducted with a calibration curve. We
know whether it is ID or OD based on which, what quadrant
the signal fall from, from 0 to 30 percent, it is ID. From
30 percent on up, it is an OD.
Well, this is similar to the second to last
viewgraph. Stress corrosion cracking depths less than 50
throughwall, we find that is not reliable, and it is not
unexpected. Smaller flaws give small signals and they are
complicated by other conditions, and it is difficult to
select the proper face angles. But what you find in general
is that these flaws are overestimated, but we see they are
unreliable because that is not always the case, sometimes
they are underestimated.
Well, and the orientation we have found is quite
difficult. Circumferential cracks at the top of the tube
sheet, when they are small cracks, they are really difficult
for the teams to get a good sizing on.
This is all that I had. I had prepared to, again,
give you a view of the work that is in progress and the kind
of data we were looking forward to getting. It will be
quite useful in evaluating submittals that come in and
getting a feeling for what the real probability of detection
of these flaws is.
MR. STROSNIDER: This is Jack Strosnider. I don't
know if you had any additional questions, but thanks, Joe.
This is some really useful work which I think is
going to help NRR in terms of our review of licensing
amendments and activities that come in. And I think the
industry, and I mentioned earlier this sort of simplified
approach to operational assessments, but I don't want to
give the wrong impression, I think licensees may actually be
out there doing these calculations and this is going to help
them do their work. I was talking about the sort of sanity
check that we give those evaluations.
In terms of schedule, we are almost finished with
Item 10. Actually, under Item 10-G, Number 1, it talks
about laboratory studies and why these are applicable in
light of vibrations induced by blowdown, et cetera. We
interpreted that as wanting to hear some more about the
issue that Mr. Spence discussed yesterday with regard to
blowdown effects.
And Jack Rosenthal from the Office of Research is
here. You know, this issue has been -- Research has been
asked to take a look at it in terms of the GSI process, and
so to address that issue, we are going to ask Jack to give
you a little status on where that is at. And I don't know
how much he has got, but when we finish that, that will
conclude Item 10, and then we can decide how to go forward I
guess with the rest of the agenda.
DR. POWERS: Well, I will tell you what the
decision there is. We will take a little break after Jack
and then we will trudge right ahead.
MR. STROSNIDER: Okay.
MR. ROSENTHAL: Really, my comments are
programmatic and short, so I will go fast. Okay. My name
is Jack Rosenthal, I am the Branch Chief of the Regulatory
Effectiveness Assessment and Human Factor Branch in the
Office of Research, and one of the teams in my branch is
responsible for working generic issues.
Yesterday that was some discussion that we have
been slow about working some generic issues, and that is
true. But since 1981, we have approached 632 issues,
prioritized them, et cetera, 283 of them actually were
worked as generic issues with some sort of technical
approach to them.
At least of recent, I think we are doing much
better at working the issues. So from 600 issues, of
course, the tough ones that take years, we resolved five in
'99, six in Fiscal 2000. There is seven on the books right
now. Our real viable process has new issues coming in and
old ones getting resolved, but the big backlog of prior
years is no longer.
The ACRS has been kind to us, and you will see we
have, with some regularity, been coming forward to you with
the issues as we resolve each technically. And you are
familiar with these because we have discussed these with you
recently.
Before us now is -- that is the list of current
issues, and monthly we tell Pete Domenici where we stand on
resolving issues. There is some thrust to get them
resolved. When this slide was made up, we didn't have
GI-188, which is the most recent one.
I was going to talk about 163, but looking through
the material, Jack Strosnider did a good timeline review at
the beginning of the day, so I won't do it again. You will
find all the information on the NRC web, and there is a
commitment there that following this panel's deliberations,
we will figure out what to do in terms of a program plan for
resolution of the multiple tube rupture issue. And in that
document sitting on the web page, it says, within a month of
you finishing your work, we will come up with a plan to
finish ours.
The last slide is 188, it is resonance vibrations
of steam generators tubes in a main steamline break event.
That is just a title that has been given to it. It has been
entered into our system, and we are starting to work the
issue. And I just -- as I understand, and this is what
needs to be worked out, the postulate is that, and it is not
surprising, if you have a fluid system and you suddenly open
that system, or you suddenly close that system, yes, one
would have pressure pulses in the system which would induce
mechanical motion in the system, et cetera. That is really
not a surprise.
At least a preliminary look, would the vibrations
be -- or the number of fatigue cycles that you put on it be
bounded by the current design of the steam generators?
Well, it is not so obvious because just the amplitude might
be different. It is something where we can't dismiss it out
of hand. It does appear to warrant some technical work.
We are following the pilot application management
directive 6.4, which we brought before the ACRS, and the
ACRS was very kind to us in retrospect when you said, look,
why don't you try it out for a year before you adopt it.
And as is proven, we have had some lessons learned from
that, so thank you.
In that process, --
DR. POWERS: We will help you, Jack. Come back to
us again and we will still harass you.
MR. ROSENTHAL: In that process, the big change is
the management directive, is to say that for issues that we
current worked, resolve really means resolved as somebody in
the street would understand the term "resolved."
DR. POWERS: Best move you can possibly make.
MR. ROSENTHAL: Okay. And that is that you bring
it through, figure out what you want to do and actually do
it and figure out, okay. And some of the activities upfront
are handled by RES, and then some of the issues, I mean NRR
has responsibility for writing rules, doing inspections, et
cetera, so it is a joint effort. But what we have said is
in terms of the public, that resolved ought to be mean
resolved all the way through verification.
We are in the identification stage of this issue.
Initial screening. We have a panel of experts, Milos
Chochki is the panel chairman on that, and the next meeting
is scheduled for 10/18.
What we do is, based on what we have heard here,
and the information that is brought forward, we are going to
try to write down what we think is the issue, get agreement
on the issue. And then -- not so simple. Because we want
to know upfront whether we are going to include things like
motion of the lowest support plate or not. Are we only
talking about the tubes? Are we talking about other
mechanical aspects of the steam generator? And just what is
the issue?
And what we have learned from other go-rounds is
that defining the issue is quintessential. Then within the
sense of the Generic Issue process, you decide if it is a
compliance issue, it was already covered by the regulations.
Is it an adequate safety issue? Typically associated with
what you think of as this 3 to the minus 3 delta CDF type
issues, which I don't perceive this to be. Or is it a
safety enhancement issue?
Following that, we would then develop a program
plan for how we would attack the issue, and then everything
the NRC does goes through a PBM process and we would get
resources to work the issues. That is where we stand.
DR. POWERS: I mean that is great, and I am glad
to see that the process is being exercised, and we will be
anxious to hear how it comes out. But we are left with a
problem now. We have a contention that says, gee, when you
set up this Generic Letter 95-05, you guys didn't take into
account the fact that you are going to get these violent
pressure pulses and vibrations in here that could lead to a
couple of things, growth of cracks that otherwise wouldn't
have grown, and enhanced leakage, and unplugging of cracks
that have been plugged by corrosion products, okay, and that
would give you enhanced leakage. So your leakage estimates
that you had in mind when you set up Generic Letter 95-05
just don't take into account this physical phenomena.
And the question is, what is the response to that?
Now, what we heard from Ken is he says the cracks are very
tight, and there isn't much in those things, so the
unplugging cracks may be not such a major issue as it is
other contexts. But the growth of cracks due to the violent
vibrations is still, I think, an open issue here.
MR. STROSNIDER: You are looking for a response.
DR. POWERS: Yes.
MR. STROSNIDER: This is Jack Strosnider. I guess
the answer to that is, number one, I think, yeah, we do need
to do some work to understand what this phenomena is. I
don't think there is anybody here right now that say how
significant it is or isn't, you know, what it is going to
do, and it is just going to require some technical work to
go figure it out.
You know, we have emerging issues in regulatory
space all the time, all right, and that is why these
processes are set up to deal with them.
The other point I would make is that, based on
what I heard yesterday, and some of the concerns that have
been expressed, it is not clear to me that this is just a
Generic Letter 95-05 issue. You know, some of the
suggestions with regard to the significance of this
transient, you know, if some of what we heard is, in fact,
what we find out when we go look at the technical aspect of
this, it is broader that voltage based repair criteria, it
has some much more fundamental issues.
DR. POWERS: That's fine. But right now I want to
work on what it has to do with alternate repair criteria.
MR. STROSNIDER: Yeah, and I think, you know, my
response is, like I say, we have emerging issues that come
all the time in regulatory space. We have a process for
dealing with them. When we talk about going out and
changing the licensing basis for plants, et cetera, we need
to do that, you know, in a methodological way, and that is
what the process is there for.
MR. ROSENTHAL: Can I make a comment? Let me just
make one more comment and then I will give you the mike.
And that is that, depending on what goes on with this panel,
okay, we have options to incorporate the resonance issue in
with 163 into a major -- into one big issue. We could parse
it out amongst its pieces. And we just haven't made a
decision pending hearing out the results of this work, plus
the panel meeting to discuss in greater depth that technical
work. And then we just have to put together.
But we are dismissing the issue. And as we look
at it, we see that there is interesting technical aspects.
DR. HOPENFELD: Let me relate to you my 40 years
of experience in Research. You don't look, you don't find.
Ten years ago, nine years ago, the broad spectrum of
problems were really identified. We didn't go into the
detail in that GSI-163 in the DPO, but NRR chose not to look
at it, chose to set it aside.
And now you tell me -- I am positive that
somebody, that if that work had started then, all these
problems would have been identified. So I am kind of a
little bit frustrated in you telling me that this is a new
thing that you are discovering today. That is water over
the bridge. The point is that with this kind of attitude, I
think we should start this new vibration program.
But if you are going to proceed in the same way
that we have done before, there will be other things here,
because it is a very complex problem. You ask yourself --
DR. POWERS: That deals with issues of management
and whatnot that are out of our spectrum. I think we are
interested in the technical issues here, and how it impacts.
MR. STROSNIDER: I would provide two additional
comments, too. I mean this is -- obviously, you as the
special subcommittee have been tasked with dealing with the
issue, and so this is just my perspective, okay, and take it
for what it is worth. I think, you know, there is a
question, and Dr. Hopenfeld just pointed to it, you know, is
this issue, was it part of the original DPO or not? And you
can take a look at that, you know, it is open to the
discussion probably or debate.
But the more important thing is, and I tried to
talk about this this morning in terms of what it takes to
resolve a DPO, or any other, you know, emerging issue and
how we deal with them, okay, saying that, you know, that in
an ideal situation you come up with "the" technical answer.
You know, we would all like to have a lot more information
on what transpired after the event down there as described
yesterday and, you know, all sorts of analyses, and we could
look at them today and say this is the answer. You know, we
don't have that.
The resolution to many DPOs, if you go back and
look at it is to say, we are going to go. You know, we
acknowledge that it is an appropriate issue for further
study and that is what we are going to do.
So that is just my perspective on it. You know,
you as a committee have to decide how you want to deal with
that.
MR. HOLAHAN: Let me just add something. This is
Gary Holahan. When an issue is referred to the Generic
Issue Program, in effect, you have made a judgment already
that you don't need to take immediate regulatory action. I
know, you know, Jack referred to the judgment about this as
concern associated with not a very high probability event,
and I think that is part of the judgment.
And I think, Jack didn't mention it, but part of
his panel's responsibility as they get into the issue is, in
fact, to identify for themselves whether this is an
immediate safety problem which could be kicked back into the
regulatory process for a Bulletin or a Generic Letter, or
calling in an Owners Group or dealing with on a more
immediate basis.
By its very nature, these things are judgmental,
because you haven't done the research work and you haven't
put all the information together, okay. But there is,
within the process, a judgment being made about this is an
issue that should be worked, and it is reasonable to take
some time to do it. And we have these sort of issues, you
know, every once in a while.
I think back to when we had problems with, you
know, fire barriers, and we talked to people, and we tried
to figure out whether that was a concern or not, and we
dealt with it a while. And then we observed the test, and
in the test, there was an immediate and obvious failure, and
two days later we wrote a Bulletin that told the industry
they had to do something in the meantime.
So when an issue moves from a concern, you know,
to a clearly known problem, we can deal with that. I think
this issue is at the concern stage. We realize that, you
know, I mean we have 2,000 years of operating experience and
we, you know, haven't had any main steamline breaks, you
know. So this issue is something that needs attention but
doesn't need attention today or this week, and can go
through a deliberate process. But as part of that process,
people have a responsibility to say this looks like more and
more like it will be resolved and it is not a problem, or it
looks like the evidence is building up that it is a real
problem. And the process has to deal with that.
DR. POWERS: I think I would have liked to seen
what the thinking about it was. Even if the outcome was
exactly what was described, are you going to put it into the
Generic Issue process?
Why don't we go ahead and take a 15 minute break.
And then we will come back, and I guess we are doing damage
propagation at that point, is that correct?
SPEAKER: Yeah, that is Item 11 on the agenda,
that's right.
[Recess.]
DR. POWERS: Let's come back into session. I
think at this time we are going to turn to the issue of
damage propagation, in particular, the subject of jet
cutting. Okay. And I have Joe and Steve listed down here.
I usually ask Steve why he is not working on the human
performance program plan, but I won't ask him this time.
So, whomsoever is leading off, please lead off.
DR. MUSCARA: In the interest of time, I could
mention, as you said yesterday, you can all read. We just
need the viewgraphs. We would rather answer questions.
DR. POWERS: Well, to tell you the truth, this one
involves CFD calculations and whatnot, and I don't read CFD
to be honest with you.
DR. MUSCARA: I don't either, that is why we have
Steve here.
Okay. So I guess we are going to be talking about
the agenda items 11 and 14, damage propagation actually.
Item 14 will be done tomorrow morning. To do this section
of the agenda, we have Steve Arndt, Steve Long and Bill
Shack will be contributing parts of the presentation.
Quickly, I will talk a little bit about our jet
impingement work that we have planned or are in the midst
of. Jet velocities --
SPEAKER: Is your mike turned on?
DR. MUSCARA: Thank you. Jet velocity and
particle motion, Steve will cover, Steve Arndt. The
quarter-inch -- the basis for the quarter-inch crack, Steve
Long will talk about that. He is here. Good. And as I
mentioned, Bill Shack will present work on different models
for predicting behavior of cracked tubes under different
conditions.
Well, the issue of the jet cutting was brought up
in NUREG-1570. I think at this point the staff really had
some concern, a lot of it based on some samples we had seen.
One of our staff members had done some work at a fossil
plant, and he was doing a failure analysis of some tubes
that had seen some jet cutting in a fossil plant, and very
impressive tubes. In fact, they did cut through -- these
tubes, I believe they are stainless steel, they are about
.44 inches think, and a jet from this fossil plant did cut
through a number of tubes. So there was some concern there.
Well, based on this experience, the staff looked
for some data that they could try to relate to the behavior
of steam generator tubes under severe accident conditions,
found some data on coal gasification and used this data to
come up with some estimates. In fact, if you look at the
NUREG, some fairly high ablation rates were estimated with
that work, where the ablation or erosion is due to
mechanical processes or corrosion processes, or a
combination of these two.
In particle droplet impingement, generally, we are
looking at a mechanism that is driven by mechanical
processes, either by jet cutting or by fatigue of the
surface layers.
For particulates in a corrosive atmosphere, the
removal mechanism can be either by the mechanical methods or
by corrosion. At low velocities, normally it is driven by
the corrosion. Intermediate velocity is a combination of
the two. And at high velocities, even in a corrosive
atmosphere, the ablation is driven by the mechanical
processes.
So we were interested in looking at this issue
again to try and relate it more closely to the conditions
that we have during severe accidents. I guess I should
mention, we will address both erosion under severe accident
conditions and under steam line break conditions, but we
need to separate those two. Certainly, under severe
accident conditions, we are dealing with high temperature
superheated steam. A compressible fluid in the steamline
break, we are dealing with water droplets and possibly
steam.
So we are going to separate those two. We are
planning some work on the severe accident conditions. That
work is underway. We are planning work also under the steam
line break conditions, that is just in the planning stages.
So to address this a bit further, we decided to do
a number of things. One was to do a literature search and
the second step was to bring together a group of experts to
talk about the issues that might be involved, in particular
with respect to the severe accident conditions and the
ablation expected under those conditions, and also to talk
about, say, the leak rates or, in particular, the creep
crack opening of the steam generator tubes under the creep
conditions we might experience in the severe accidents.
So we held a specialists meeting at Argonne
National Lab on November 19th, '99. We do put minutes
together and those are available to the public. They were
sent to the Public Document Room on December 10th. At the
meeting, it was open to the public, but we particularly
invited a selected number of experts. Among those in the
erosion area, we had Ian Wright from Oak Ridge National Lab
and John Stringer from EPRI. I guess I should also mention
that the data that was used in the NUREG from the coal
gasification work was data developed by Ian Wright, among
some others, but that was a major part of the data that was
used.
In the severe accident area we had Jason Schaperow
from NRC and Mati Merilo from EPRI. In the high temperature
fracture mechanics we had Professor Saxenna from Georgia
Tech, and then the various other from NRC and ANL, including
staff from Combustion and B&W, or ABB and FDI.
When we discussed these issues, certainly a number
of things came up as being important in this area, and a
number of clarifications were provided by the experts. Both
Stringer and Wright felt very strongly that the fossil
experience with the superheated tube could not be used or
extrapolated to the steam generator case.
In particular, the fireside atmosphere in a fossil
plant contains a heavy load of ash particles, sand and the
large particle sizes. And what happened in the cutting
there is that the jet entrains these very abrasive
particles, their large size, and they cause the cutting.
So, you know, we don't think this kind of particle is really
present in the generator.
DR. POWERS: Numerous speakers have spoken of
sludge piles and whatnot.
DR. MUSCARA: Yes.
DR. POWERS: The oxide itself is spinel. It seems
to me that there are some fairly hard particles in there.
DR. MUSCARA: Yeah, we actually discussed this
with the experts, you know, the possibility of the jet
picking up particles as it exits the tube. Well, there are
several locations in the generator where this could be
possible. One place in particular where you have sludge
probably would be the top of the tube sheet. Another place
might be at the support plate. I don't think we get a great
deal of sludge there, you know, not as much as we get at top
of tube sheet.
The consensus was that, if you know the nature of
the sludge, there is sludge lancing that goes on
periodically to get rid of the sludge at the top of the tube
sheet, and the loose particles are usually taken away by the
sludge lancing. But, in fact, the majority of the sludge is
not even able to be lanced off. The stuff is cementatious
and it is very hard and sticks to the tubes. So we thought
even if the jet worked its way through a piece of sludge, it
may pick up a few particles, it would tunnel through there
and would not really pick up much more beyond that.
DR. KRESS: Under severe accident conditions that
generated a lot of aerosols.
DR. MUSCARA: Yes, I will get to that, sure.
DR. KRESS: You are going to get to that later.
DR. MUSCARA: Sure. Yeah.
DR. POWERS: Under design basis accidents, won't
you be carrying in lots of the crud particles from the
primary piping system in the jet?
DR. MUSCARA: Yes, generally you do get corrosion
of the carbon steel area. I mean much of the primary system
is clad, but there is some carbon steel, you do get some
product. We have done some work in the past trying to
characterize leak rates through cracked pipes, and, you
know, we try to do a search and get information on the kinds
of crud that you get from the primary side. It is really
not crud. You may have some very small particles, and even
if you have those, the loading isn't that great.
So, you know, we are not -- we haven't quantified
that. Our feeling is that you do not have a large amount of
crud due to the corrosion products that gets carried by the
primary side fluid.
Also, Stringer felt that the droplet erosion
during design basis accident was unlikely, and the reasoning
is that the erosion rate is dependent on the droplet size,
and it is related to the diameter to the third power. We
have noted water droplet erosion in steam turbine, but this
occurs because of fine droplets condensed in the turbine,
and when they enter the turbine, they become larger drops
and then the spinning blades hit these drops and you get
erosion, which the insiders call baseball bat erosion. But,
again, you know, these are large droplets that the finer
droplets are condensed and then are picked up by the blades.
DR. CATTON: So do the smaller droplets cause more
problems?
DR. MUSCARA: No, less. The big droplets to the
third power.
The NUREG-1570, extrapolation of the data from the
coal gasification plants, they assume that the ablation rate
will be proportional to the density of the fluid and to the
cube of the velocity. The temperature affected the
extrapolation only as it changed the density of the fluid.
But, in effect, the work that was done for coal
gasification, the gas mixture is very oxidizing. In
particular, it is 1 percent H2S in this mixture, and nickel
alloys under corrosion in these kinds of atmospheres.
So what we are looking in the coal gasification
data is one at done at high temperatures, much higher than
we expect, and at higher temperatures the corrosion, you get
a greater amount of corrosion. In addition, the work was
done at low velocities, 10 to -- I believe they had a set of
data at 10 feet per second and a set of data at 100 feet per
second. And then, of course, this was extrapolated up to
about 1,000 feet per second, using the correlation to the
third power.
DR. CATTON: Can I go back to that first paragraph
for a moment? What is the scenario that you are looking at?
Isn't it the high pressure inside the tubes and its water?
DR. MUSCARA: Under steamline conditions, yes.
DR. CATTON: And isn't that what we are talking
about?
DR. MUSCARA: Yes.
DR. CATTON: So the jet would expand from, I don't
know what, 2,500 psi down to 1,000? Isn't this what leads
to the erosion, so it is the droplet sizes associated with
the fragmenting jet, liquid jet?
DR. MUSCARA: Right.
DR. CATTON: So where does this fine droplet
business come from?
DR. MUSCARA: In the turbine case, the water
droplets coalesce, become large droplets.
DR. CATTON: But here you are starting with a
liquid jet.
DR. MUSCARA: Right.
DR. CATTON: And it is going to fragment into
small droplets.
DR. MUSCARA: Right. So they are small --
DR. CATTON: How fine are the droplets?
DR. MUSCARA: Right.
DR. CATTON: They can be coarse.
DR. MUSCARA: We will be addressing the area of
the jet behavior later. But let me just mention right now,
we --
DR. SHACK: This is just somebody's opinion, an
opinion.
DR. KRESS: It is experts telling what they think.
DR. CATTON: Okay.
DR. MUSCARA: In the literature. But let me
just --
DR. CATTON: At the agency, we know quite a bit
about this kind of process because this is what is
associated with combustion. So you don't have to think it,
you could base it on something that is real.
DR. MUSCARA: Right. Right now the first step was
to concentrate on the severe accident condition.
DR. CATTON: Okay. It just seemed you were
throwing that one away.
DR. MUSCARA: Right. I am bringing this up also
because we will be doing some work in this area. So my
feeling is we need to understand the dependencies of the jet
and how it expands, the particle size, particle density, et
cetera, for the severe accident case because we can't really
conduct tests under those conditions.
We are also trying to understand how the droplet
erosion would work under steamline break conditions, and we
will try to understand that from the literature as much as
we can. However, we have developed a very nice facility at
Argonne National Laboratory for conducting tests under
prototypical conditions. So regardless of what the theory
tells us, my first step is to conduct -- well, they are
concurrent steps, but we are conducting actual test under
prototypic conditions with cracks, and whatever jets that
are produced impinging on a sample. So there we get some
data on the prototypic conditions.
Meanwhile, we will also try to understand it from
a theoretical basis.
For the severe accident case, there is no way that
we can develop a rig to produce the kinds of conditions that
you get under severe accidents. So here we are depending
more on whatever knowledge is there, what other research has
been done. So this is the one I would like to address
first.
DR. HOPENFELD: If you want, I will make my
comment later, but it is pertinent to this point, if it is
okay with you. I will make it very fast.
Three years before the DPO, I asked Los Alamos to
do a study as to what happens when a jet 2200 flushes into
water and flushes into steam, into air. To come up with
some kind of estimate, what kind of particles, particle size
you have. They have done a very considerable amount of work
on that. The conclusion was, with all due respect to the
expert, that you cannot really predict what size you have.
You can come up with sizes from one micron all the way to a
fraction of a millimeter.
So what I am kind of seeing here, that you are
starting a new program without really looking at what
happens based on a meeting. Now that is not how you do
research.
DR. MUSCARA: So when one extrapolates the coal
gasification data, this is data really that is based on
corrosion, not ablation, and the dependency to the third
power doesn't hold. In fact, when you look at the data
itself, work was done at different velocities, different
temperatures, it is inconsistent with their extrapolation.
And, also, the effect of temperature on corrosion was
ignored.
Based on the literature review, and the experts
meeting, we identified some of the key parameters. Two of
these were the jet velocities and the associated particle
motion that were some of the most important parameters,
including the particle size.
Having this background, we asked for some
assistance from our Division of System Analysis and
Regulatory Effectiveness to carry out calculations to better
define the jet velocities and the particle motion that we
would expect under severe accident conditions. This work
has been completed, and I think I would like to break at
this point and ask Steve Arndt to address some of the
findings from this work. Steve.
MR. ARNDT: Thank you. As Joe mentioned, my name
is Steve Arndt. This week, literally, I am the Assistant
Branch Chief for the Safety Margins and Systems Analysis
Branch in the Office of Research. I am going to go through
some of the work fairly quickly that we were asked by the
Division of Engineering to look at to support some of their
work in characterizing the kinds of damage propagation you
can get.
There have been several attempts to come up with
an appropriate velocity impinging on the adjacent tube to
this kind of damage mechanism. We were asked to look at
this for basically three primary reasons. One, to get a
better fundamental understanding of what is going on. The
previous analysis were fairly simple analysis. Two, to
understand what the particles within the fluid were doing,
which had not been looked at previously, because the
computational tools weren't available. And, also, to
support the work that the Division of Engineering is going
to be doing at the University of Cincinnati, or is actually
in the process of doing at the University of Cincinnati, and
to benchmark the velocities that they need to test at.
And when we originally discussed it, one of the
things was, is 1,000 feet per second an acceptable number?
Will it give you the numbers?
So, as you can see, this work was done by
Professor Piomelli, who is a professor of mechanical
engineering at the University of Maryland who does CFD
calculations. The code that we used was the NPARK code. I
believe Professor Catton is familiar with that. It is an
Air Force developed code that is specifically for high
velocity flows.
Because we were trying to understand the
phenomenon and also provide input to the Division of
Engineering, we wanted to look at, based on our
computational study, what not only the important parameters
were and the actual numbers, but what the sensitivities
were. So we looked at variations in the temperature, the
pressure, the various steam geometries and the crack
thickness. I will show you that in a moment.
For this particular study, we did a
two-dimensional study, so we have a crack, we looked at two
different thicknesses and assumed an infinitely long crack.
The particle size and densities were developed from a
Victoria calculation, and this is a slightly misleading
phraseology, Charlie pointed this out to me, we assume an
equal distribution in the tube at the time of the crack. It
is not in the entire primary system, it is the particular
density at the primary side of the crack. And we also
assumed that the particle velocities were calculated along
with -- I'm sorry. One of the goals was to calculate the
particle velocities.
DR. POWERS: So you are looking really at the
severe accident scenario?
MR. ARNDT: Yes, this is a support of the severe
accident scenario.
DR. KRESS: Are these Victoria calculations, were
they using the natural convection recirculating
countercurrent flow conditions?
MR. ARNDT: That's correct.
DR. KRESS: They actually used those.
MR. ARNDT: We used the SKDEP calculation to
develop the accident scenario and then used the Victoria to
actually propagate the aerosols.
DR. KRESS: So these aerosols went back and forth
with some residence time that may be relatively long while
you are heating up and agglomerating perhaps and changing
size, and you got all that out of Victoria?
MR. ARNDT: Charlie.
MR. TINKLER: Yes. Charlie Tinkler from the NRC
staff. As it turns out, most of the larger aerosols are out
of the stream by the time they get to the steam generator.
DR. KRESS: They fall out --
MR. TINKLER: They fall out. So we are left with
a distribution, I think it was 1 to 5, on the order of 1 to
5 microns, something like that.
MR. ARNDT: Yeah, I will show the distribution
here. Actually, I will show it now just because we are
talking about it.
MR. TINKLER: You know, we had a large, a
relatively large inventory of non-radioactive aerosols that
were floating through the system.
DR. KRESS: That was going to be my next question.
What was your source term for non-radioactive aerosols?
MR. TINKLER: I think we have three or four
hundred kilograms worth.
DR. KRESS: Coming from the cladding?
MR. TINKLER: Coming from the cladding and
structural materials in the core. I think I actually show
the number in one of my viewgraphs in tomorrow's
presentation. I think it is about three or four hundred.
DR. HOPENFELD: Can I just make one comment?
Because what I said yesterday, you cannot do those because
of the agglomeration that you get, because the
thermophoresis forces in the plenum. So unless you have, in
the mixing plenum, unless you have the temperature
distribution, you can't figure out the residence time.
Right now they don't have it, it is a perfect mixing.
DR. KRESS: Do you include thermal phoresis at all
in the calculation?
MR. TINKLER: Yes. Yeah, we have thermophoresis
in the model, and, typically, in the tubes themselves,
thermophoresis is relatively small effect because the
temperature difference between the vapor and the thin steam
generator tubes is pretty small. But in other parts of the
system, you know, we use the SKDEP RELAP boundary
conditions, you know, through a rather tedious process of
imposing as boundary conditions on the Victoria. We use
their thermal hydraulic conditions, and, yes, we do have
thermophoresis.
MR. ARNDT: This is the relative particle size,
out of particle mass, particle size and density of the
distribution that we had at the time of the opening. And,
as you can see, the mean is in this couple of micron area.
You might want to remember this because, for
various reasons associated with the computation, we refer
the mass as opposed to the particle size, which is not quite
as intuitive in later calculations. But I will keep that
handy in case anyone needs it.
Like I said, we looked at a variation of several
different things, temperature, size of the crack -- I'm
sorry, size of the crack which is this -- these are actually
half-heights, because we use a symmetric system. Two
different sizes of a rectangular grid and a triangular grid.
These were designed to be similar in configuration to a
Model 51 D type steam generator and a triangular grid from a
combustion engineering.
And what you have here, and, by the way, all the
details are in this handout that you got. I believe you now
have the color version of this, with all the gory details.
What you see, and these are the physical properties. Let me
actually skip to the next couple of slides later, because it
is a little easier to explain what is going on. This is two
slides later. From the velocity curves as opposed to the
thermal properties curves.
What you have is a thin hole, a small hone, very
high pressure, low pressure. This is 16 megapascals, this
is atmospheric, so you have an expansion pressure ratio of
160, which produces a very under-expanded jet.
You go through rapid pressure drop and velocity
increase and you, because you have a blunt body here, form a
shockwave. The expansion would have continued out into, if
you didn't have this actual blunt body here, to a typical,
very high, under-expanded jet kind of phenomena. We did
some sensitivity studies basically by removing that and
looking at what would have happened had it not been there
and got pretty much what you would expect from a theoretical
standpoint.
If these had been parallel plates, you would
expect it to expand, then drop, the pressure would drop
rapidly and then as you go over here, would increase almost
to the original pressure. The same for the velocity, you
would start it at a lower velocity, expand rapidly to a very
high velocity, the mock numbers are very high, as you can
see, and then drop to almost zero here.
DR. KRESS: I presume these are steadystate
calculations where you kept the primary side pressure
conditions constant?
MR. ARNDT: Yes, that's correct.
DR. KRESS: Okay.
MR. ARNDT: Given the flow rates and the
availability of steam on this side, basically, you can't
deplete this in the kind of timeframes we are talking about.
And this sets up very quickly. You are on the order of a
couple of microseconds to set up that kind of steadystate.
Because both of these barriers are curving away,
the flow does not stagnate here, but actually shoots off in
this direction. In the rectangular grid, of course, there
is another tube up here, and you would have another
shockwave up here. In the triangular grid, there is another
tube over here, and, basically, what you have is a second
nozzle type effect where you are shooting off fluid off this
direction and off this direction.
If you go back to the original graph and look at
the thermal properties, you can see, after the shockwave,
well, the jet shoots out very high density, expands rapidly.
After the shockwave, the pressure goes back up, and the
density goes back up, but not nearly as high as the original
pressure. You go through a very rapid drop in temperature
as well on the other side of the shock. Because you are
compressing the fluid, you increase temperature again. And
you can see that from this particular graph. The density
drops down dramatically, the temperature drops and then goes
back up again. This is the velocity in the X direction and
the pressure.
DR. CATTON: You approach velocities of about 3600
feet per second in that. You have 1200 meters per second,
about 3600 feet per second.
MR. ARNDT: Yes. And if you look at the
variations associated with the different geometries, you can
see the basic phenomenology is very similar. You expand
rapidly, the pressure drops, the velocity goes up. You go
through the shock front, the velocities drop, the pressure
goes up.
DR. KRESS: Where is the 1,000 feet per second on
that?
MR. ARNDT: This is -- 1,000 feet per second, it
would be right in here.
DR. KRESS: So that is where the number comes
from?
MR. ARNDT: Well, I will show you where that comes
from.
DR. KRESS: Okay.
DR. CATTON: Normally, you don't get such nice
shocks when you use this code. Did he do something to
smooth them?
MR. ARNDT: We looked at --
DR. CATTON: You get more spikes, unless you do
special effects.
MR. ARNDT: Yeah. These are all center line
calculations, along the --
DR. KRESS: Along the line of symmetry.
MR. ARNDT: Along the line of symmetry. We will
look at in a minute what it looks like off line of symmetry,
and you will see it is a little more --
DR. CATTON: Anybody who gets such beautiful
shocks, I don't trust it. And I have used NPARK.
DR. KRESS: A shocking statement.
DR. CATTON: Usually you get wiggles, you get the
wiggles just because the codes can't really treat the shock
that well. You have to do special effects in order to treat
the shock.
DR. KRESS: Well, he had an extremely fine grid.
MR. ARNDT: Yeah.
DR. KRESS: And he had very, very small time
plates.
DR. CATTON: You get spikes.
MR. ARNDT: We also, like I mentioned, --
DR. CATTON: Magnitudes don't change a whole lot.
MR. ARNDT: -- did several sensitivities based on
things like temperature. We varied the temperature by 100
degrees, and we saw fairly small changes. We varied the
size of the hole quite a bit. We see significant amplitude
changes, particularly in the pressure, because you are
putting out a whole lot more fluid, so the pressure behind
the shockwave will be considerably higher, but the basic
phenomenology is fairly similar.
Now, of particular interest is what is happening
along the streamlines, and that is because one of the things
we are really interested in is what is happening to the
particles along this flowpath. If you look at jet cutting
tools, they are very high colonated, usually gets of water.
Reasonably high velocities, very high pressures, very high
particle loading. And they, of course, will turn when they
hit the piece of metal you are cutting to, but the braces,
as well as the actual water, will actually go forward.
So, one of the things we wanted to look at was
along the line -- not along the line of symmetry, but
actually various jet streams. The first is actually outside
of this jet, and you can see it goes through a much
different velocity and mock number profile. If you look
along these two, particularly, the third one here, which is
fairly close to the center line, but off of it and does the
turn, what you see is it, of course, accelerate, decelerates
through the shock, comes to a steadystate during its turning
here, and then as it goes up through here and starts
accelerating again into the nozzle between the two, you get
the accelerating again.
This is basically where we got our 1,000 feet per
second. Somewhere in this range will be right off or right
near the center line.
Now, if you want to look at the particle
velocities, you have to look at what the particles are doing
in the fluid. They are going to be accelerated with the
fluid based on, in essence, a relative flow rate between the
particles and the fluid, that gives you the driving force to
accelerate them. And in these particular cases, you can get
all sorts of different forces associated with them, but the
dominant one is the drag force on the particle.
And if you use the basic Reynolds number for
particles, the stand Stokes Law type calculation, what you
get is kind of what you expect, although when we did it, we
didn't think it was going to be this dramatic. The smaller
the particle, and if you will remember, our mean particle
density was right in this area, a couple of microns -- I'm
sorry, micrometers, what would happen is it would accelerate
as it goes through the expansion, then it would decelerate,
but the amount of acceleration and deceleration basically
depending upon the size because you had -- you are driving
the drag based on size. We, as it turned out, for
computational convenience, used the drag on a sphere. We
later looked at what would actually happen if it was not a
sphere, which, obviously, a lot of aerosol particles aren't,
and I will talk about that in a minute.
Of particular interest, of course, is it takes a
little longer for the heavier particles to accelerate, as
you would think. It also takes longer for them to
decelerate. Again, the particles that we were looking at,
by and large were in this range. And you see they drop off
rather dramatically. We have a few particles up in this
higher range, but this wasn't real satisfying, so we want to
look at this a little bit closer.
If you look at the data for very high Reynolds
numbers for particles -- let's see if I can find my graph
here, you find that the standard and Stokes Law doesn't
really apply very well. And use the Stokes solution for the
drag coefficient on a sphere with a relative Reynolds
number, it will predict something like this. If you go back
and look at some of the experimental data in this, it
doesn't really do that. So if you go -- what we did was,
for very low velocities, we used the Stokes solution. For
this intermediate range, we used an older solution, and then
we also used the experimental data to try and redo this.
What happens when you do that is the larger
particles, even though they do slow down -- rather, speed up
and slow down at a slower rate, they are considerably more
dependent upon the fluid velocity than if you used the
straight Stokes Law.
So the real issue here, as we wanted to really
find out, was what was the fluid velocity doing? What was
the particle velocity doing? And then we can give that to
our friends in --
DR. CATTON: Are these particles solid or what?
MR. ARNDT: They are assumed to be solid.
DR. KRESS: So your major conclusion is that the
particles are going the same velocity as the fluid.
MR. ARNDT: They are going the same velocity as
the fluid and, more importantly, they are moving with the
fluid. If you go back to the streamline analysis, they are
moving with the fluid, and they are decelerating here, and
they are also turning. If they weren't moving with the
fluid, they would have the tendency to go forward. They
would basically maintain their momentum and go forward in
this direction. So they are moving with the fluid and they
are also moving at the fluid velocity. So they do have the
tendency to turn with the fluid due to the blunt body.
DR. KRESS: The first calculation using Stokes
Law.
MR. ARNDT: Yes.
DR. KRESS: They didn't follow the fluid.
MR. ARNDT: They didn't follow the fluid as much.
They tended to not be decelerated with the fluid because the
drag coefficients were lower, so they would decelerate
slower and also turn less.
DR. KRESS: Does that imply there might be an
optimum drag coefficient?
MR. ARNDT: There probably would be. Because
aerosols are not nice perfect spheres, we also did a
sensitivity study that increased the drag coefficient and
decreased the drag coefficient by a factor of 10. And I
didn't plot it up for you because it is not in the report as
a plot, but what basically happens is, by doing that, you
move this up a little bit. It comes in kind of like that.
DR. CATTON: So what this is saying is basically
you can't erode the adjacent tube with a compressible flow.
MR. ARNDT: With this kind of compressible flow,
with these kind of particles.
DR. CATTON: Is it because of the size of the
particles?
MR. ARNDT: That is primarily --
DR. CATTON: There are examples, in the old Nike
Zeus program, they were going to steer it with the Cunard,
and they just chewed a hole right through it. In that case
the particles were actually liquid, less penetrated.
MR. ARNDT: Yeah.
DR. CATTON: And the velocities weren't near these
because it was still, it was in the nozzle, it was still
turning. And that kind of -- maybe the particles were
bigger or smaller, I don't know.
DR. POWERS: I guess the question I have for Dr.
Kress is, what impactors work? That sonic jet is coming in
on plates?
DR. KRESS: They work because the particles --
DR. POWERS: Can't make the turn.
DR. KRESS: Yeah, can't make the turn, that's
right.
DR. POWERS: And don't stay with the flow
velocities.
DR. KRESS: That's right.
DR. POWERS: And here they are tracking, this
impactor won't work. But one micron particles, one micron
particles are the easiest particles in the world to get an
impactor to work on, because they don't stay with the stream
velocities at sonic levels.
DR. KRESS: Flow velocities through an impactor
are smaller.
DR. POWERS: No, they can be sonic, but just
sonic.
MR. ARNDT: Now, if this didn't --
DR. CATTON: The shock standoff distance seems to
be quite large also.
MR. ARNDT: Yeah, it is. If, for example, this
was at sonic, or subsonic, you wouldn't have this kind of
shock standoff. Sonic velocities are down in this range for
this particular fluid at this particular temperature. Also,
you have comparatively low particle loading in this
particular case, which is something that we are -- I think
Bill is going to talk about a little later.
The sonic velocity is down in this range. So if
you assume that you have very near sonic velocity, say, for
example, up in the next tube over, say, for example, this
will expand and contract. Then you are down here, but you
are also at a much lower density not of the fluid, but of
the particles themselves.
DR. KRESS: What do you mean lower density of the
particles?
MR. ARNDT: Well, as you spray this out, you are
expanding the jet.
DR. KRESS: Oh, you are talking about number
density.
MR. ARNDT: Number density, yes. I'm sorry.
DR. POWERS: I also know its impact better when
the number density is lower.
MR. ARNDT: Yeah. This is the study we did and
these are the conclusions we came up with.
DR. HOPENFELD: I would just like to make a quick
comment.
MR. ARNDT: Yes, sir.
DR. POWERS: Is this an item of verification?
DR. HOPENFELD: No, it is just a clarification,
that is all. For two phase flow, in subsonic flow, if you
have gas and particles, the loading of the particles is an
important factor because there is an interaction there, they
tend to stack up. That affects the standoff distance and
their loading rate. And I was just wondering, I am just
making a suggestion, so I am not criticizing anybody here,
maybe you should look also, look at it as a two-phase flow,
set up the basic equation for it, and this way you can find
out what the effect of concentration is, just like a
two-phase flow equation basically, particles and gas.
DR. KRESS: I suspect your number density of these
particles is so small, you are not going to affect the sonic
velocity in this case. It is a pretty small number density
there, so it looks pretty -- it is mostly acting like a gas.
MR. ARNDT: And, of course, if you have any
additional questions, I am available tomorrow, and, of
course, we can provide additional input.
MR. HIGGINS: You didn't have a slide on the
conclusions, but you said verbally that the conclusion from
that was that they would not, the particles would not erode
the other tube. That wasn't really the objective of this
part of the study. Part of the study objective was to
provide particle and velocity -- particle and fluid velocity
calculations to the Division of Engineering so they can look
at what particles moving at those kind of velocities, in
those kind of densities would do, and they are going to talk
about that next.
DR. KRESS: The particles that are right on the
line of symmetry have nowhere to go except impact the tube
-- so those at least will go on and impact.
MR. ARNDT: Yes. And we would expect them to
impact in the two or three hundred meters per second kind of
time velocity.
DR. KRESS: I guess the question may boil down to,
do those particular particles do some sort of damage to the
next tube?
MR. ARNDT: Right. And Joe is going to talk about
some of the experimental work he is doing on velocity,
particles of that velocity and that size on actual pieces of
zinc alloy.
DR. MUSCARA: Okay. So, based on the information
we have been gathering, we have set up some tests, tests to
address the jet impingement under severe accident conditions
to be conducted at the University of Cincinnati with
Professor Tabakoff, and we are also in the process of
planning and running some tests at Argonne National
Laboratory under steamline break conditions.
The conditions we are considering for the severe
accident conditions, a temperature of 700 degrees centigrade
and a pressure 2350 psi. The particle loadings, we
discussed earlier, taken from the code. You see most of the
particles are silver, about 85 percent of the particles are
silver. There are some oxides, 10 oxides dominant and
indium oxide.
The total loading is 115 grams per cubic meter.
The medium --
DR. POWERS: It is a little surprising you don't
have any urania in that mix.
DR. MUSCARA: Any?
DR. POWERS: Urania.
DR. MUSCARA: You know, I suspect it has to do
with its melting and volatilization temperatures. If it is
not volatilized, it doesn't get picked up in the stream and
condensed later on into an aerosol.
Charlie.
MR. TINKLER: I am not sure we are claiming it is
zero. I think we have just listed the dominant species and
compounds that might be present. I doubt very seriously if
it was zero UO2 in there.
DR. KRESS: It looks an awfully lot like it is at
the stage of the accident where you have just failed the
control rods and attacked the plant a little bit, but
haven't gotten --
MR. TINKLER: That is also true. You know, in
past presentations, I have indicated that typically these
are the conditions at the time that we normally predict the
surge line or hotleg to fail.
DR. KRESS: Okay. So you haven't really --
MR. TINKLER: They are still relatively early in
the core degradation, overall rows core degradation process.
DR. KRESS: You have probably entered the high
rate of steam zirc reaction.
MR. TINKLER: Yes.
DR. KRESS: Just barely probably.
MR. TINKLER: We are into the temperature
escalation of the cladding and the core, but we haven't
gotten to the formation of a large molten pool or things
like that yet.
DR. MUSCARA: That is why we try about 700 degrees
centigrade, it reflects the temperature of the tubes at the
time of surge line failure, which is at 684 degrees under
the 6 RU scenario that is described in many of Charlie's
reports.
DR. SIEBER: Once the surge line fails, the
driving force can make the jet --
DR. MUSCARA: This is why we are concentrating
here, yeah.
DR. SIEBER: So that is a reasonable assumption.
DR. MUSCARA: So the mean particle diameter is 1.5
microns, most of the particles were less than 3 microns. I
think the distribution I saw, there might have been a few
particles at 5, but nothing at all beyond 5.
So before we are planning this work, we looked for
places where you could conduct some experiments. And it
turns out there are a couple of rigs around the country. At
the University of Cincinnati, there is this apparatus that
has been used for many years and many erosion studies
conducted by Professor Tabakoff. We decided that this was a
place we could conduct some experiments.
This is the rig that is used. Essentially, there
is a propane burner atop of the rig. Air is mixed with the
fuel. Below this there is a preheater for the particles, so
the particles are fed out to the preheater, with some time
in residence to pick up temperature. And then the particles
are injected in the stream and there is a fairly long
tunnel, acceleration tunnel. And at the bottom, here is
where the test specimen is. There is a capability for
changing the angle of the specimen with respect to the
fluid.
Beyond that, there is an exhaust tank where
effectively these gases cool, the particles can drop out and
recover.
Well, the atmosphere certainly is not pressurized
steam, but it is an oxidizing atmosphere. Most of the
combustion products would be CO2 and steam.
DR. KRESS: What particles do you use?
DR. MUSCARA: Well, that is one of the following
viewgraphs. The predominant particles that we had in the
aerosol, as mentioned, was silver. And we can't use silver
for these tests, in particular because the silver would melt
in the combuster. So we were looking for a surrogate
material we could use for silver and still be conservative.
So what is important here, of course, is the density and the
size of the particles, and essentially the hardness of the
particles at that temperature.
So when we compared different materials to the
major particles in the aerosol, we settled on using nickel
for the majority of the particles and to simulate the oxides
with nickel oxide. And, also, to be even more conservative,
we were looking at some aluminum oxide in conjunction with
the nickel.
We did need information on the velocities,
however, when we run these experiments, we like to go beyond
the particular velocity that was calculated, just to make
sure that we have enough information. So we run tests from
a lower velocity, about 300 feet per second, up to 1800 feet
per second. The initial series of tests were aimed at
determining the worst conditions, so we ran a number of
tests at 1,000 feet per second by changing the angle of the
target. We looked at, I believe, 20, 30 and 45 degrees, and
the maximum wear rates were obtained at 30 degrees. This is
similar to many other tests that have been conducted. This
is a similar angle that produces the worst results.
So the subsequent tests were run at 30 degrees and
we varied the velocity and also the particle mix.
Again, I must say that these tests are not
complete, but we have completed tests with the nickel
powders at 300, 600, 1,000, 1,800 feet per second, 100
percent nickel. The particle size, 3 to 7 microns. And
there is the initial data on the erosion rates.
We are also planning on running tests with nickel
plus 15 percent nickel oxide and nickel plus 15 percent
aluminum oxide, and have completed some tests with the 100
percent nickel oxides, thinking that this would be as
conservative as one could get, all of the particles are the
hard oxide.
DR. POWERS: Your data seemed to indicate that
something unusual happens between 1,000 and 1,800 feet per
second.
DR. MUSCARA: Well, there is the relation of the
wear with respect to velocity. Professor Tabakoff had some
tests some years ago using 70 micron quartz and, clearly, he
gets higher wear rates, but the velocity dependence is
similar, and it is representative of what happens with
mechanical abrasion.
DR. KRESS: Now, the units on this are cubic
centimeter per gram per second.
DR. MUSCARA: Yeah, actually, those are converted
numbers. What we get from Professor Tabakoff from the tests
are milligrams of material lost per milligram or gram of
particles impacting the surface. So we converted those
numbers and are taking into account the temperatures. We
convert to get an estimate of the amount, depth of material
that is worn.
And for the 300 meter -- well, 1,000 feet per
second, and with the nickel powder, the wear rate on this
material is about 4 mils per hour.
Now, I am not sure if I have a viewgraph on this,
but the experiments we have conducted with the nickel oxide,
in fact, we did not get anywhere. We effectively had
deposition. There is the sample weight more at the end than
it did before. But at these temperatures, of course, the
material is fairly soft, and the hard, abrasive particles
embed themselves into the material. They plow the material,
they don't cut the material, so we had -- we didn't have any
wear there. So now we are going back and trying the nickel
with 15 percent nickel oxide and also nickel with 15 percent
aluminum oxide just to see whether there is synergistic
effect there, but I suspect limiting data will be the data
with just the nickel powder.
You know, I have indicated that we will be running
some tests for the steam line conditions, but in some of our
prior work for different purposes, we are running some tests
on different size orifices. We did have one test where one
32nd inch hole was impinging on the inconnel target. This
was a test at room temperature, with 2500 psi pressure, a
four hour duration test. At the end of the test we noticed
only very light burnishing of the tube.
We will run tests under more prototypic
conditions. Well, first, conservative tests at 2500 psi and
300 degrees C, and then following those tests, we will
consider running some tests that more closely reproduce the
accident scenario. There is the actual pressures and time
relationships.
DR. POWERS: And these are liquid tests that you
are talking about?
DR. MUSCARA: This would be tests where we
effectively, on the primary side, we have high temperature
water. The secondary side will be dry and we will see what
happens.
Well, I think quickly to conclude, we believe the
damage by jet impingement, due to severe accident conditions
in particular, is not going to be a concern. We are doing
additional tests, I think that will be confirmed.
I guess, also, we conclude for the rest of this
section, which has to do with the work that Bill will talk
about tomorrow, but effectively develop models for
predicting the structural behavior of a good tube, as well
as degraded tubes under normal operating design basis
accident conditions and severe accident conditions. We will
see all of that tomorrow. But the models have been very
good in predicting behavior and we have quite a bit of test
data to validate those models.
So I think at this point, I am finished unless
there are some questions.
DR. POWERS: Any other questions on this
presentation?
I don't know these results are all stunning -- I
mean surprising, are they, to you? I mean, typically, when
you use jets to cut things, you work with much higher
velocities?
DR. MUSCARA: Much higher loading on the particles
and much larger particle sizes. Our meeting with the
experts, they really felt there would not be much erosion
under those conditions. We felt that there was enough
evidence from the literature and from the meeting that that
would be the case, but we wanted to run some tests to verify
it.
DR. CATTON: I think the use of a CFD code like
NPARK, which is really, it is a code that has been around
for a long time, it was originally developed by Los Alamos
and then the Air Force picked it up, and they actually have
an office in St. Louis whose only purpose in life is to
incorporate everybody's experience. It is a reliable code.
The only question one might have is the shock standoff
distance might not be quite right. But looking at the
particle sizes, they are correct, and the drag coefficient
is anywhere near correct, the tracking of the particles to
the fluid velocity, you could have that shock standoff
distance and it still wouldn't cause much of a problem. I
think they did a fairly substantial job in demonstrating
that this particular aspect is not a problem.
DR. POWERS: My concern remains the same, it is
counter-intuitive to me to have particles like that tracking
the velocity so closely. Because usually we rely on the
fact that they don't track velocity closely in order to
sample and trap them.
DR. CATTON: That's right. That's right.
DR. POWERS: But I mean I have no experience that
at these kinds of velocities.
DR. CATTON: My only experience was the opposite,
and I cut a hole right through the device, but it was a lot
hotter.
DR. KRESS: I would follow up a little on Dana's
comment. The code you are talking about doesn't have
particles in it, it calculates the stream lines.
DR. CATTON: It is pure and simple a compressible
flow code.
DR. KRESS: The question I have then is how did
one translate the drag coefficients in order to see whether
the particles followed the stream lines or not.
DR. CATTON: Well, you have to make an assumption
when you do this. The assumption is that the particle
density is low enough that it doesn't impact the flow.
DR. KRESS: That is a reasonable assumption.
DR. CATTON: If the loading starts to go up.
DR. KRESS: No, that is a reasonable assumption on
his loading we have here.
DR. CATTON: If you can make that assumption, then
it is just a matter of fitting the particles into the flow
field and asking, where do they go?
DR. KRESS: Do you have the capability of putting
them in the flow field and changing the drag coefficient?
DR. CATTON: Not with NPARK, no.
DR. KRESS: As you move from one spot to the
other.
DR. CATTON: I suspect what they did is they just
have a data set with a velocity field in it, and they stuck
the particle in and said, where do you go? Is that correct?
SPEAKER: Yeah.
DR. CATTON: Yeah. So once they have the velocity
temperature and pressure fields.
DR. KRESS: So all you have to do is put one
particle of each size at a given spot and watch it go?
DR. CATTON: That's correct. That's correct. And
then just look at its trajectory.
DR. POWERS: I mean the place where we run into
problems with that is so much higher that I mean I don't
have any trouble with these assumptions. I have troubles
with the conclusion because I mean, how do we move particles
across boundary layers? They don't track fluid velocities.
And especially when you get up to a micron, I mean it is a
micron where we have impaction problems. You get much below
a micron, then, yeah, they track the stream velocities
pretty well. Certainly, if you get down as low as a tenth
of a micron, then they really track stream velocities well.
DR. CATTON: What are the drag coefficients?
These are --
DR. POWERS: It is a drag curve, a lot like the
one he showed with the dots that come across there. I mean
the drag coefficients you had were about -- seemed all very
rational to me. So I am perplexed, I mean my experience in
shockwaves is exactly zero.
DR. MUSCARA: I had almost forgotten, but there is
another item on the agenda, and that is to discuss the
behavior, the reasoning for the selection of the
quarter-inch crack.
DR. POWERS: Right.
DR. MUSCARA: Steve.
MR. LONG: After what just transpired, I think I
need to remember everybody that I am about to talk about
what happened first.
DR. POWERS: In the beginning there were
quarter-inch cracks, right.
DR. KRESS: Your first test is to see if you can
turn the thing on.
DR. POWERS: If you can't turn that thing on, then
we are not going to believe a word you say.
DR. KRESS: It is the big long bar on the front.
MR. LONG: There we go.
DR. POWERS: Now we have to listen to him.
MR. LONG: Okay. To put this in context, when we
did NUREG-1477, we were dealing with really just
measurements of cracks in terms of voltage, and we didn't
have length and depth information. We were somewhat
concerned about severe accident issues where there would be
high temperature flowthrough cracks in tubes. We didn't
really -- I'm sorry. We didn't really have any way of
working with that until we got some distributions of crack
sizes. So, through a research contract, Dominion
Engineering produced some correlations of crack sizes and we
could proceed when we did NUREG-1570 work.
DR. POWERS: Correlations of crack sizes with?
MR. LONG: Well, basically, they were giving us
results of a lot of different analyses that the utility
companies had done with their distributions in their plants,
separated out by distribution -- by degradation type, and
then trying to give us a distribution of lengths and a
distribution of depths.
DR. POWERS: This is just the database they
maintain on what people find.
MR. LONG: I don't know if they maintained it or
produced it, but we received it through the subcontract.
They didn't correlate the length and the depth, and when
they did the correlations to the depths with the data, they
used gamma functions that basically fit the data in the
exponential part of the function and had offscale low in
depth and in length, extremely high artifact peaks that
didn't fit any of the data.
So the difficulty was we ended up, when we tried
to combine the two to get a distribution of physical cracks
estimated, we would end up with a very large number of
extremely short but extremely deep cracks.
Well, when we looked at the temperatures and what
we expected the cracks to behave like at the temperatures
the tubes would reach before RCS pressure boundary failure
in core damage accidents, the type of behavior we were
seeing would have said a crack that was maybe 4/10ths of an
inch long would be about as short as you would expect to
rupture with a typical limit load analysis based on the flow
stress, as best we could extrapolate it.
So, we were aware of the DPO concern about
cutting. I think there was also some consideration of
cutting when NUREG-1150, accident progression expert
elicitation was done. So we wanted to try to represent the
cutting somehow assuming the cracks that were still too
short to rupture would open significantly and might be able
to cut adjacent tubes, or for that matter, erode the hole in
that tube that the flow was going through.
But the problem was we had a large number of
cracks that could be throughwall that were perhaps less than
a hundredth of the inch long in the correlation, and if I
just put that in there, I had a probable certainty that
those cracks would be present. Should I assume they cut?
Talking to the materials people, this was
seemingly resolved with sort of the classical back of the
envelope argument that stress corrosion cracking had
typically as aspect ratio that was at least five times
length to depth and, therefore, for a 50 mil tube, we really
shouldn't see throughwall things much shorter than a quarter
of an inch.
That sort of became the cut-off in NUREG-1570 for
looking at the distributions that were given to us by
Dominion.
That picture sort of stayed in vogue until we
really started doing profiles of cracks that were coming
from plus point analyses of flaws found in power plants.
And the first one I was afflicted with was from Farley when
they requested essentially a waiver of a mid-cycle
inspection on the last fuel cycle before they replaced their
steam generator tubes. And the way this data was being
treated is that the eddy current signals were being treated
as planar cracks with jagged shapes, and then these were
being projected over the cycle to grow, at least in depth,
and I have forgotten if they were growing theirs in length.
Sometimes they are not grown in length, sometimes they are,
depending on who is doing the analysis.
At any rate, once they are grown in a Monte Carlo
process by depth, they are analyzed for the fraction of the
crack that might go throughwall and perhaps create a leak or
go throughwall and burst. And this is done by
mathematically taking a rectangle, taking a small length and
moving it along the crack, taking the average depth within
that length, calculating the stress magnification factor,
then taking a slightly longer length and doing the same
thing until the find the part of the crack that has the
maximum stress magnification factor for pop-through and the
maximum for burst.
What Farley found, very late in the review
process, actually, Westinghouse was doing the analysis, was
that they had a near certainty that they would have
something go throughwall by growth during that cycle. The
concern was that they thought it was very short, and did we
want to treat that as a complete failure of the primary to
secondary boundary for the risk analysis?
This was a difficulty because now we weren't
talking about a whole crack that went throughwall, it was
sort of a simplified rectangular approximation to the crack
shape. Now, we were talking about long cracks, they might
be half an inch, an inch long, but we were only talking
about a small segment, so we couldn't argue they don't
exist, they probably do exist.
So we had to figure out, did we believe a crack of
a certain length that was throughwall would really open
significantly and leak significantly? We didn't really have
data that would allow us to do this back when we were doing
the Farley analysis.
So it ended up with a telephone conversation late
in the game. Bill Shack was on it, Joe Muscara was on it.
I am sure Bob Keating was on it. I think Tom Pitterly was
on it, but I have forgotten. I was, some people from the
Farley site, and we were trying to figure out what we would
do in this case. Would we basically reject the application
on the idea that some crack, not matter how short, would
penetrate the wall during the fuel cycle? Or would we stick
with the quarter-inch or set some other value?
And I think the general feeling was, from the
people that had some experimental evidence but hadn't really
been able to quantify it, so it was a qualitative feeling,
that the quarter-inch crack was probably not going to be
severely cutting, not in the timeframe that we thought we
had before some other part of the pressure boundary would
fail, if, in fact, this didn't fail first and relieve
pressure.
We didn't really know what number to pick. It
wasn't a quantified judgment. So at this point what we
decided to do was leave the crack at the quarter-inch as the
threshold for what we would consider to be a gross failure
and proceed with the application review.
We wrote the SER and pointed out that the results
would be sensitive to this conclusion. I will go into the
practices of risk-informed decision-making on my slides
tomorrow, so at this point I will just say that one of the
principles is to look at the things that your decision is
sensitive to and the uncertainties.
We put in that this was a judgment, and I know
that that upset the DPO author because he felt should have
something better to make a relaxation. The best we could do
at the time was to make that clear, that it was not an
analytical result, it was a judgment, that we were sensitive
to that judgment, and to modify the DPO considerations
document as well so it was clear in there that this was an
issue that we needed more information on.
Subsequently to that, RES started the work that
you just heard about. That was actually before we formally
got the letter over to them to request them to do that. And
as you have heard, at least from the cutting angle, it
doesn't seem as though there is a real problem with the
quarter-inch cracks. And there is another aspect of this
that has to do with how much those cracks would open up and
leak, and Joe Donohue is not here today to present that, so
I guess we will have to deal with part of that tomorrow.
There is still then, now we are talking about
quarter-inch openings that might exist, there is still the
potential issue about how much leakage would you get through
them. And there what we would like to do is make sure that
we are at least staying below the 1 GPM that is the current
tech spec limit.
In that regard, you would want to have, you know,
very few cracks. And another part of the work that I didn't
hear about was to try to add the creep aspect of the crack
opening, so that you could get a crack area and get a more
valid calculation of leak rate tomorrow.
I hope that explains the history of not answering
all the technical questions.
DR. POWERS: Okay. Any questions on this? We
have a quarter-inch cut-off for cutting, which may not occur
at all, but we still, we don't have any cut-off for leakage?
MR. LONG: The leakage has not been handled
quantitatively at this point for very small cracks, in terms
of looking at a distribution and quantifying it through
that.
DR. POWERS: Any other comments that the members
would like to make?
[No response.]
DR. POWERS: Well, on that note, I will thank
everyone for some very nice presentations, very informative,
and invite you all to reappear at about 8:30 tomorrow for
some more of this fun.
[Whereupon, at 7:00 p.m., the meeting was
recessed, to reconvene at 8:30 a.m., Friday, October 13,
2000.]
Page Last Reviewed/Updated Tuesday, July 12, 2016