ACRS Meeting on the Ad Hoc Subcommittee on Differing Professional Opinion Issues - October 13, 2000
UNITED STATES
NUCLEAR REGULATORY COMMISSION
***
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS
***
AD HOC SUBCOMMITTEE ON DIFFERING
PROFESSIONAL OPINION ISSUES
***
Friday, October 13, 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
CHAIRMAN POWERS: The meeting will now come to
order.
This is the fourth day of the meeting of the ad
hoc ACRS Subcommittee on Differing Professional Opinion
issues. The purpose of the meeting is for the subcommittee
to review technical issues contained in the differing
professional opinion on steam generator tube integrity.
The subcommittee will continue to hear from the
NRC Staff today. In particular, we will continue our
discussions of damage propagation, then we'll hear specific
discussions on design basis accidents, severe accidents and
integrated decision-making.
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 this
meeting. Ms. Undine Shoop will be around someplace to
assist us.
We have received no written comments or requests
for time to make oral statements from members of the public.
A transcript of the 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 members of the panel or the consultants have
any comments to make before we return to the general
discussion of damage propagation? They all look glassy-eyed
today. I think you did 'em in yesterday. They are not
feeling too frisky this morning, I can tell. The first
speaker is really lucky.
On my agenda I have Joe continuing.
MR. MUSCARA: Thank you and good morning.
I do not have too much to say this morning, but
we'll continue with the damage propagation with a
presentation from Dr. Shack on integrity of steam generator
tubes.
CHAIRMAN POWERS: You will have to tell us more
about his background and why he is qualified.
[Laughter.]
DR. KRESS: Is he qualified to speak to us?
MR. MUSCARA: If I were you, I wouldn't listen to
him.
CHAIRMAN POWERS: We don't in any case.
MR. MUSCARA: There is one comment I would like to
follow up from the last presentation yesterday.
I started yesterday talking about POD and the fact
that we need a robust POD test -- do not really come up to
100 percent, even for large flaws, and the reason I gave of
course was the human factor, but the human factor also
affects small flaws as well as large flaws. If the person
is blinking, whatever --
Another reason we don't always get 100 percent POD
for eddy current inspections of course is that the voltage
can be quite low. I am not sure how rare an event this is,
but you can get a flaw that is long and deep and provide you
a very small response.
In fact, I indicated yesterday the mockup contains
on the order of hundreds of tubes. Within these hundreds of
tubes we had about four flaws that were not detected by any
of the inspectors, the reason being they were small voltage
flaws, I believe below one volt.
In fact, these flaws are large. They are on the
order of up to two inches long, 80 percent deep or deeper,
so you can miss large flaws not only because of the human
factor but unfortunately they have low voltage response.
CHAIRMAN POWERS: On the first day of our
discussions of these phenomena we had something of an
explanation of why you would get that. I mean it was a
plausibility or intuitive description that they have lots of
these cross-ligaments in a tight flaw so they remain good
conductivity paths. Is that your understanding as well?
MR. MUSCARA: Precisely. This is my understanding
and speculation at this point.
However, whenever we have a test result like this
where it's been detected and we believe there are flaws
there and we have sized them with our own techniques we will
take the specimens out of the mockup and section them and
verify what the condition is and what might be causing the
low response, so our inspections indicated these are large
flaws.
I mean clearly we can see the length -- and by the
other methods we believe they are deep but we will section
some of those flaws to make sure that indeed they are deep
and whether there are a lot of ligaments with these.
MR. CATTON: How well does the search for flaws
perform when you look in the vicinity of the support plates,
or is this a part of this study?
MR. MUSCARA: In the mockups?
MR. CATTON: Yes.
MR. MUSCARA: Yes, the support plate is an area of
interest. There are flaws there and the techniques are
being used in the field for that kind of flaw being used in
the round-robin.
MR. CATTON: What is the POD?
MR. MUSCARA: Well, like I say, we are just in the
midst of pulling the data together and deciding what true
state for the generator is and conducting these evaluations.
At this point I really don't have the answer, but
we cannot have all this work, not all of it finished but
much of it finished so that we can have a topical report by
the end of this calendar year and at that point we may very
well have that information.
MR. CATTON: So it will be a part of what you do?
MR. MUSCARA: Oh, yes. In fact, we have included
with the support plate the complicating factors of the
crevices being blocked up but also denting, with assimilated
denting and superimposed flaws on the denting, so we made
the test reasonable enough that it represents the field
condition.
I think unless there are some other questions I
would like to have Dr. Shack come up and talk about the
integrity models.
DR. SHACK: Just for the record, I am Bill Shack
from Argonne National Laboratory and I qualified to speak on
this subject mostly because I have a bunch of competent
colleagues at Argonne who do the work.
Let me just hand out some toys. This is a steam
generator tube unflawed, pressurized to 2350 and taken to
840C, so this is what happens in the high dry sequence to
the unflawed tube if you get the temperature high enough.
It would be bigger except that we have a two inch
guard tube around this thing and so when it opens up and
smashes into the guard tube it kind of flattens out.
MR. CATTON: It becomes bubble gum.
DR. SHACK: I passed around a sample stress
corrosion crack yesterday that I thought everybody -- I
assume everybody found the 360 degree circumferential crack.
You might just want to compare what that looked
like in your memory with an EDM notch so you can see how
good a simulation an EDM notch is of a stress corrosion
crack.
Let me pass this one around after I do a little
bit of talking here.
I am just going to briefly review a lot of work
that was done by the NRC and industries during the '70s and
'80s to develop verified models for failures of flawed steam
generator tubes and I am going over that because some of the
work that we did at Argonne was to extend that work to the
short deep flaws. That was one sort of shortcoming in some
of the work that was done at PNL. They didn't have enough
short deep flaws, and so we wanted to go back and to extend
the model and do a little more testing just to see how we
were doing with the short deep flaws.
Then into this sort of traditional failure of
tubes under design basis conditions we have, as you have
heard, got into an extended sort of discussion of the
potential failure of steam generator tubes under severe
accident conditions, in particular under these high dry
scenarios where we have a depressurized secondary side and
then the core melting just drives the temperatures up to 700
or in fact if nothing else in the system fails to even
higher temperatures.
As I have noted, flawed tubes will fail at lower
temperatures but even unflawed tubes under the sort of 2350
pressure if you take them to 800 to 840C will fail rather
spectacularly.
Again as we get up to about 700C you can see the
flow strength of Alloy 600 is decreasing markedly and again
what we find of course is that in these tests, as you would
expect the failures at low strain rates are controlled by
creep and at high strain rates we expect them to be
controlled by flow stress.
We have typically found that we do better with
creep failure models, so we, whenever we can and we have
enough data, we try to work with the creep models.
For a part-through crack now let me say that
failure can mean two things. Failure can mean -- in a part
through crack means I have a crack that is, say,
three-quarters of an inch long but it is not all the way
through the wall. It's, say, 80 percent through wall.
Well, I can have the failure of the ligament when the crack
pops through the wall to create a leaking crack and that can
occur in two ways. That ligament can pop through but the
length of the crack doesn't change, so that is a stable
failure and what I end up with is a leak, and if I have a
three-quarter inch crack that pops through I have a leaking
three-quarter inch crack.
Unstable failure or burst means that the crack not
only pops through but will grow in length without any
substantial increase in pressure or burst, and what I have
here is a sample of a tube. It's sort of an interesting
one. It is a three-quarter inch long EDM notch, which is
sort of nice because you can make everything exactly
precise. It is 80 percent through-wall --
MR. CATTON: What kind of crack can you get with
the EDM?
DR. SHACK: It's about three to four mils.
MR. CATTON: And the walls are very smooth?
DR. SHACK: Yes. This is wide open. This is a
hole.
MR. CATTON: That's what I thought.
DR. SHACK: But from a structural point of view it
is a crack. That is, this is not a ceramic -- from Alloy
600 an EDM notch and a stress corrosion crack are the same
structurally. There's a difference certainly in leak rate
because the one is far tighter than the other, but as far as
the structural behavior goes, an EDM notch is a very good
simulation.
What you will see on this crack, you will see kind
of a bright, shiny line at the bottom. That is the
remaining 20 percent ligament, and if you look real
carefully you will see it slides off at a 45 degree angle.
It is a sheer lip failure -- that thing popped through.
Then you will see some tearing at the ends of the
crack, again on 45 degree lines as this thing is stablely
tearing. What happened is this one popped through at 2850
so we got the 20 percent ligament to fail at 2850 but it
wasn't an unstable failure. We had to go to 3000 psi before
we got the tearing at the ends of the crack and as it began
to extend in length.
Now again, what happens exactly when you start to
get the unstable tearing is kind of unclear because in all
laboratory systems we run out of pressure, and of course
when we run out of pressure the system stops growing. In a
real plant, yes, you'll run out of pressure but you will rip
a big hole and again after a hole gets so big it doesn't
make a whole lot of difference.
Once this crack is about an inch and a half long,
the crack opening is about as big as the diameter of the
tube and so the flow restriction is really the tube. It is
no longer the crack, so any crack lengths over an inch and a
half are almost kind of -- you know, not terribly exciting.
This is an interesting crack in the sense that it
would not have failed, even the ligament would not have
failed under a main steam line break, if you will allow me
just to use pressures for the moment.
MR. CATTON: The mild main steam line break.
DR. SHACK: The mild main steam line break, where
we just increase the pressure to 2500. Even the ligament
would not have failed but this is a tube that wouldn't pass
the 3 delta p criterion, so this is in that in-between
range, not good enough to pass 3 delta p, but it wouldn't
have even failed the ligament under the main steam generator
break, or shall we say the depressurized secondary loading.
MR. BALLINGER: What you are saying is this stuff
is pretty tough.
DR. SHACK: This stuff is pretty tough.
The bad news about Alloy 600 is it cracks. The
good news is it's tough as hell.
A variety of models have been used to describe the
failure, unstable failure of through-wall cracks and the
ligament failure. Most of them involve this kind of stress
multiplier factor. It is really a bulging factor and you
will notice that the axial tube here fails in a bulge way.
What I want to note is that this bulging factor
depends on the radius so curvature counts here and one of
the things that we will see, and we should keep in mind, is
that tubes are much weaker to axial cracks, because we have
this R-factor, and you can sort of see if I go to a flat
plate as R gets very large, this bulging factor goes down,
down, down and in fact I should have brought the flat plate
solution and one of the interesting things about a
cylindrical tube is that it has got a curvature in the hoop
direction and it is a flat plate essentially in the axial
direction, so that in fact axial cracks under the same
stress will open up a lot more than circumferential
stresses.
If I have, for example, a quarter-inch flaw in the
axial direction it will open about six times wider than the
same quarter inch flaw in the circumferential direction
because again under pressure loading I have a two-to-one
pressure ratio and I have a multiplier of about three
because of the curvature effect for the dimensions of this
tube, so again even if I had the same loading in the axial
direction that I had on the hoop direction the hoop crack
would open up about three times as much as the axial crack.
DR. KRESS: Bill, what is the V in that equation
or the Greek letter?
DR. SHACK: Here?
DR. KRESS: Yes.
DR. SHACK: Pousson's Ratio, .3.
DR. KRESS: Pousson's Ration, okay.
CHAIRMAN POWERS: Okay, keep going. What is C?
DR. SHACK: C is the half crack length. R is the
radius of the tube and T is the thickness of the tube.
CHAIRMAN POWERS: And the bulging factor is this
M --
DR. SHACK: M.
CHAIRMAN POWERS: -- which is not dimensionless?
DR. SHACK: Yes, it is dimensionless. That lambda
is a dimensionless quantity -- C over square root of RT.
DR. KRESS: Right.
CHAIRMAN POWERS: Okay.
DR. SHACK: In Christian units inches over square
root of inches squared.
For part through-cracks, we have a similar
formulation, but we have a different expression for the
bulging factor, and we won't worry too much about that.
There is a fairly extensive database that goes
through the burst pressure and the ligament failure
pressures to validate those correlations.
And, again, you'll notice, unlike the voltage
correlations, when you go to a more mechanistic correlation,
I can put 3/4, 7/8 and 11/16ths inch tubing all in the same
plot, if I non-dimensionalize with the square root of RT,
and I non-dimensionalize the burst pressure.
DR. CATTON: So what do you think is work with the
data that we looked at yesterday? It's just not presented
correctly?
DR. SHACK: With voltages, you can't
non-dimensionalize. There's nothing wrong with it; it's
just that you'll need separate correlations for 3/4-inch,
11/16ths, and 7/8ths-tubing.
DR. CATTON: I don't understand that.
DR. BALLINGER: If you knew the crack length
exactly --
DR. CATTON: So that's the problem; I don't know
the crack length.
DR. SHACK: The problem is that you don't know the
crack length.
DR. CATTON: Okay. I don't know what's causing
the particular voltage reading, okay.
DR. SHACK: The way I like to look at these things
is sort of a geometry failure map here, and I'm looking at
what happens to the whole range of flaw geometries that I
could have in terms of the length of the crack and the depth
of the crack.
And everything below this curve, all cracks here,
will have no failure at normal operating pressures, so I can
have three-inch crack, 85 percent through the wall, and
under normal operating pressures, that crack is going to sit
there with no problems.
If I have a crack that's one inch long, it will
pop through when it gets to be a little over 90 percent
deep, so it will pop through. But it will pop through
stably; it will pop through to give me a one-inch,
through-wall crack that will not grow in length, but will
sit there and will leak at some rate, and we'll talk about
leakage later.
However, if I had a three-inch crack that got to
about 87 percent deep, it would pop through and it would
start to run unstably until -- but again, three-inch crack,
once it popped through and opened up, I'm dead anyway.
DR. KRESS: Are those lines pretty thick?
DR. SHACK: No, those lines -- Yes, I should
mention that. The lines here, think of them as about an
eighth of an inch fuzzy line will cover the range of stress
of material properties that I have in the tubing.
So draw them with a magic marker kind of thing.
DR. BALLINGER: Does that include the vertical
line?
DR. SHACK: Yes, the vertical line will also
shift, depending on how things go.
Now, on some of the plots where it has mattered,
I've sort of shown the 95/95; on the plots where I haven't
shown it, just think of fuzzy lines.
Now, if I go to a main steam line break, the
geometry picture changes a little bit. Again, I need a
crack that's something over 70 percent through-wall of any
length to fail under the main steam line break conditions.
If I have shorter cracks, again, let's take a look
at the quarter-inch crack. That has to be about 95 percent
through-wall to fail, even under a main steam line break.
So, again, if we're talking about short cracks
popping through and leaking under these conditions, we're
talking about short, very deep cracks.
Again, anything below 85, you know, I need a
fairly substantial crack if it's not going to be at least 85
percent through-wall.
DR. CATTON: When you run these tests, everything
is nice and quiet, and the tube is sitting there.
DR. SHACK: We'll talk about that.
DR. CATTON: If you shake it just a little bit?
DR. SHACK: We'll talk about that.
DR. KRESS: That's saying under one inch never has
an unstable burst?
DR. SHACK: On a main steam line break, right.
DR. BALLINGER: How is the vertical line
determined? How is the dividing line determined?
DR. SHACK: Well, I have essentially a
through-wall crack margin and a pop-through margin. When
the pop-through pressure exceeds the unstable growth
pressure, that's when I get --
DR. BALLINGER: So it's experimentally determined?
DR. SHACK: No, no. It's analytically determined,
but it's also verified. Either the curve that you showed
there, showed the burst correlation versus the ligament
failure correlation, so this is one case when I know the
geometry, I can predict the stuff, you know, quite
accurately.
CHAIRMAN POWERS: Yesterday at some point in the
discussion we had a rule of thumb about crack depth being a
fifth of the length quoted to us for -- it was for
estimation purposes.
Is there some range of validity of those kinds of
rules of thumb?
DR. SHACK: I think that was trying to estimate
the shortest crack that would go through wall, and that
doesn't strike me as an unreasonable number. In a case like
this where there is no particular microstructure, to somehow
focus the crack growth and take it through, and that really
follows almost from fracture mechanics type arguments when
you look at the kind of growth that you could get in the
length.
Now, we can get longer cracks, you know. You can
obviously get cracks that are longer than five times the
depth, but I think that's a reasonable number for short
through-wall cracks.
But again, let's look at some of the consequences
of short through-wall cracks in a little bit, after I get a
little further along.
DR. KRESS: Look at these curves, Bill, where does
the 40-percent through-wall in the rule come from?
DR. SHACK: Okay, we're just about to get there.
DR. KRESS: Oh, I'm sorry.
DR. SHACK: Because, again, this is normal
operating pressure, main steam line break. But we're
looking for a three delta-P margin, and, you know, you're
always asking what is the margin?
Well --
DR. KRESS: Here, you really know what it is.
DR. SHACK: Yes. The NRC has determined that
three delta-P is it. You know, we go no lower. And the
answer, of course, is that an unflawed tube has a margin
that's probably nine times delta-P.
And you've allowed that to decay, but the margin
-- you know, you've said that it will go no lower than three
delta-P.
And you will notice that three delta-P, now, we
had sort of a 60 percent based on wastage, but again, you
get about the same number for a long crack. A short crack
can obviously tolerate a much deeper kind of thing, so,
again, short, deep flaws are not a problem.
DR. KRESS: But you go ahead and assume there's
long cracks?
DR. SHACK: Yes, but if you're going to assume
there's a long crack, then the 40-percent through, so, you
know, if you were -- if you were changing your 40 percent
rule, you might -- and if you thought you could predict the
crack depth, and you thought you could predict the crack
growth, then you might, in fact, do it based more on this
whole overall curve.
But, again, they've kind of argued that, you know,
you take a kind of an average, a worst-case kind of thing,
and you'll end up with the 40-percent through-wall.
DR. KRESS: But still this is 65 percent.
DR. SHACK: Yes.
DR. KRESS: It's not 40.
DR. SHACK: My guess is that they calculated the
60 percent based on a code minimum yield stress for Alloy
600.
DR. KRESS: I see.
DR. SHACK: I calculate -- Westinghouse did a very
nice job collecting yield and low stress data on all the
heats of Alloy 600 out there, and so I'm using sort of 95/95
and mean stress values on those kinds of yield stresses,
rather than code minimum, so, you know, a regulator may well
use a code minimum, but I'm a researcher, so I'm allowed to
be more realistic.
That's all very nice, but in many ways, we're
leak-rate-limited. You know, if you look at those curves,
you need big mother flaws to fail unstably, so in many ways,
it's leak rates that control these things.
And so what I've shown here is a crack opening
area versus crack length. And you can sort of see that
things start to get exciting here under normal operation
conditions when you get out to about an inch, and they get
very exciting when you get out to about an inch and a
quarter.
And, again, you begin to see a significant effect
of yield strength on the crack opening area that you get
from the longer cracks. And for reference here, I've sort
of shown the crack opening that corresponds to when you just
sliced the tube off and you've got the ID area in relation
to this crack.
And this curve is just this curve on a log scale
so you can really see what's happening down here in this
short crack range.
And, again, how do we calculate these? Well, we
calculate them from linear elastic fracture mechanics. We
use a particular model, due to Zahoor.
We've done essentially finite element analyses to
verify this model; we've done tests where we do essentially
room temperature leak tests so we can get a flow through a
crack and, you know, measure the area of the crack that way,
and then compare it with what we predict from the model.
We've take pictures of these cracks, scanned them,
digitized them, taken pictures of them.
And, of course, like all fracture mechanics
models, they're better, the smaller the deformation. You
know, these are all small deformation models, so that the
smaller the opening, the better.
But it is remarkable how well it does. We had one
of these little sort of freebie jobs we did for the Swiss.
They wanted some ruptured tubes. They were going to use
them for a test.
And, of course, being the Swiss, they didn't ask
for ruptured tubes; they wanted ruptured tubes with an
aspect ration of the crack opening to the crack length, and
they specified it.
So, you know, you're sitting here with a curve
that's about to go vertical, and you're trying to hit the --
DR. KRESS: You're trying to stop on that aspect.
DR. SHACK: You're trying to stop on the dime to
match the Swiss thing, and, of course, you know it -- but
the amazing thing is, that even for these rather large
openings, we were able to do a very good job at predicting
them from our model, and so we supplied designer ruptures to
the Swiss.
DR. KRESS: Now, when you do a finite element
around a crack like that, you have to get very small?
DR. SHACK: Yes, I can do the Zahoor analysis in
an Excel spreadsheet, you know, and the calculation takes
one blink of an eye.
DR. KRESS: But in your finite element, does the
crack end up at a short vortex?
DR. SHACK: No, it will round off.
DR. KRESS: It rounds off?
DR. SHACK: Yes, especially in these.
VOICE: [Off microphone.]
DR. SHACK: To look at the flow through these
cracks, there are a couple of things of interest, so
obviously the first thing if interest is the crack opening
area. That tells you how big the hole is.
But the other thing I want to know, is what's the
L over H? And, again, a lot of work was done on this for
stress corrosion cracks in piping, but stress corrosion
cracks in steam generator tubes are a little different,
because sometimes they look like holes, and sometimes they
look like long thin tubes.
So, if I've got a short crack, I've got an L over
H, depending on whether I'm in main steam line break of
something over a thousand, or, you know, 500, so I'm looking
down a very long narrow tube.
If I've got a crack that's more like half an inch
or three quarters of an inch, I've basically got a hole.
And so you get sort of different fluid mechanics models from
that.
The other thing that's interesting to look at --
and this is a plot that is not in your book, but it was
handed out as a separate viewgraph today -- and that's the L
over D for a leaking jet.
And, again, one of the things that's of interest
when you have a jet, of course, is the diameter of the jet
versus the distance it has to go before it impacts the
target.
And so we if we look at the L over D for a jet of
dimension .125 inches, since we had some concern about
cutting from steam jets of cracks of 1.25 inches or smaller,
we see for a 1.25 inch crack, the L over D is 2000.
DR. KRESS: What are you talking about here?
DR. SHACK: The .25 inches to the next tube
divided by the diameter of the exit jet.
DR. KRESS: Okay. That's just geometry.
DR. SHACK: Just geometry, just geometry, but it's
an important geometrical parameter to keep in mind.
DR. KRESS: Okay.
DR. SHACK: So for a .125 inch crack, it's 2000,
if I look that the L over D. Just as a point of reference,
the CFD calculations you were looking at yesterday were for
an L over D of eight.
DR. CATTON: Was it because they picked a really
big hole?
DR. SHACK: Yes. They're fluid mechanics guys,
and they don't know how big a crack opens up. They picked a
.5 millimeters that seemed like a good idea at the time.
Then they doubled it to 2.5.
DR. KRESS: These were rectangular holes. Is the
D there just the width of the --
DR. SHACK: We're talking slots here. Even a
quarter inch crack is an infinite slot when you look at the
crack opening here.
DR. KRESS: So when you say D, that's just the
width of?
DR. SHACK: The width of the opening, right.
Coming back to this crack opening area, let's just
talk a little bit about leak rates through these cracks.
We mentioned a model called Crack Flow that
Westinghouse uses. One of the simple-minded calculations is
just a simple pressure drop, you know, orifice model.
And the nice thing about that is, it gives you a
bounding leak rate. So if I take the full delta-P and I
divide it by rho, take the square root of two times that,
times .6, I get an orifice flow equation.
And if I apply that to .125-inch crack, I find I'm
leaking .03 gpm. So I'm not sending a whole lot of liquid
out of this crack, and, of course, it gets smaller at a
fairly rapid rate for cracks less than .125.
Now, in fact, of course, since my L over H ratio,
which I have shown here on this plot, even under a main
steam line break for this .125 inch crack, is about three or
four hundred. There is, in fact, a significant pressure
drop.
DR. CATTON: With pressure ratios like that,
shouldn't you use compressible flow equations?
DR. SHACK: Yes, but the non-compressible flow is
a conservative estimate, so my .03 gpm is a conservative
estimate. I'm just -- there is a variety of models. We
talked about Crack Flow, and, again, a lot of work has been
done on this in connection with stress corrosion cracking.
There's a model -- you know, Westinghouse has
Crack Flow, the NRC has Squirt. Professor Schrock has a
code called Source. EPRI has a code called PICEP, and
PICEP, Squirt, and Crack Flow use the -- again, you have to
do compressible flow models here.
And as Dr. Hopenfeld mentioned, there's a
non-equilibrium thing. There's a -- you start out as
liquid, and they flash to steam, but, in fact, you can get a
metastable state where the flashing doesn't occur when you
-- and it's not a thermodynamic equilibrium at all times.
And PICEP and Crack Flow use the Henry model for
discussing that transition from the non-equilibrium
situation to the equilibrium situation.
Professor Schrock has a different model that he
developed for the NRC. The code is called Source. There's
a NUREG on it.
He's done a fair amount of careful experimental
work, and I can leave it with the Committee, if they are
interested. But I think the important conclusion from
Schrock's experiments is that when you use the Henry model,
which is what Crack Flow uses, you're conservative.
And he says you can be conservative up to an order
of magnitude for the geometries that Schrock examined.
CHAIRMAN POWERS: Conservative with respect to?
DR. SHACK: Predicting mass flow through the
crack.
DR. KRESS: You predict more than you get?
DR. SHACK: You predict more than you would get.
So, you know, again, I haven't been through Crack
Flow to find out whether they do the sums correctly, but,
again, based on Schrock's evaluation of it, a model using
the Henry correlation for discussing the transition from
equilibrium or non-equilibrium transition, is going to give
you conservative results for the critical mass flow rate.
DR. CATTON: And Schrock is very careful.
DR. SHACK: Schrock is very careful.
DR. KRESS: When you say equilibrium, I'm not sure
I understand what you mean. I think you're talking about
metastable state, right.
DR. SHACK: It's not really equilibrium.
DR. CATTON: They talk about there's two; you can
talk about frozen flow, which means whatever the fluid is,
it stays at the inlet side conditions.
Or you talk about equilibrium flow. There it
thermodynamically adjusts at each stage along the flow path.
Or non-equilibrium flow where you can be -- the flow can go
further down the -- it goes down the hole and is not in
equilibrium with its pressure.
And that becomes a much more difficult kind of
calculation.
DR. SHACK: Yes, and Schrock does a true
metastable thing where's got basically a time constant.
DR. CATTON: What people should normally do is,
you do frozen flow, do equilibrium flow; and if they are not
too far apart, you quit.
DR. SHACK: Okay, one of my conclusions from this
is, because of the L over D ratio and the low mass flow
through the .125 inch crack and the very large L over D for
this thing, is that you're very unlikely to get steam jet
cutting from these short cracks.
The Argonne tests will be done for cracks that are
-- for geometries that are more characteristic of a .4 to .5
inch crack for which the L over D ratio is much smaller, and
you get much more mass flow through the crack.
DR. CATTON: How small? Is it still on the order
of 100?
DR. SHACK: What, L over D?
DR. CATTON: Yes.
DR. SHACK: Yes.
DR. CATTON: A hundred is still low.
DR. SHACK: We're going to run the tests.
DR. HOPENFELD: Will you give me one minute?
DR. SHACK: Sure.
DR. HOPENFELD: Because this is a very subtle
point, and I don't think I was really describing it at the
time because of the time that I had. I didn't get into the
detail, but this is an opportune time to express the point
exactly why is it important about whether it's one phase,
frozen flow, or whatever.
If you go back to your proprietary data, you see
that there was this extrapolation or equation, using the
equation of pressure and temperature, very simple square
root type equations.
I don't want to talk about it, but it's in your
data there. And there is a question, evidently, that people
who came up with those equations wanted to make sure that
they can justify that.
So, what they did, they went back to -- now I can
say what code it is. It's Crack Flow.
They went back there, and they used the voltage
data to come up with some kind of effective length, because
for the crack flow you need the length of the crack, right?
So they came up with some kind of a crack length as a
function of voltage, and then they plugged that thing back
and they got a line comparing that theoretical prediction of
crack with the database.
And they say, ah, well, that's fine; that looks
very good. Okay, and therefore we are confident in the
database that it has some theoretical justification.
And my point at the time was, wait a minute; you
can't say that, because you don't know whether you had a
two-phased flow in those tests or whether you had a
one-phased flow or what you had, because in that crack flow
you don't have the ten to the minus four metastability.
And there was the point, you know, is that
obviously can forget all the two-phase flow, and you would
be conservative, just the way you're doing it, and just
using an orifice equation.
And that probably is the way to do it, but my
point was that they are trying to justify that all that
database --
DR. SHACK: But my point is that because Crack
Flow uses the Henry Model, Schrock's results -- and, again,
I'll be glad to donate my coffee-stained copy of Amos and
Schrock to the panel, if they'd like to look at it -- it
says that those results will be conservative.
DR. HOPENFELD: I'm not questioning the
conservatism; I'm just trying to bring the point that the
line that the drew to compare with the database doesn't
really prove anything.
It doesn't get you a better feeling that they know
how to extrapolate from the laboratory test where the
pressure was not the same, to the steam line break; that's
my point. I'm not hundred percent right, that it's more
conservative.
DR. SHACK: No, the laboratory test was run. I am
almost positive that the laboratory test was run at the
right pressure. It might well have been run at a lower
temperature, because, again, it's a lot easier -- I can run
room temperature tests at 2500 psi without any difficulty.
Running tests at 2500 psi and 300 C is a more
expensive thing, so I suspect they were correcting for the
temperature.
DR. HOPENFELD: No. There was the delta-P wasn't
the same. The back pressure was not the same. I suggest
you go back there and take a look at it.
If it wasn't proprietary, I probably would have
picked up those points, but I suggest you go back there and
read all of that. There is a lot of material there.
And you'll find out now that that's why they have
all these corrections. And some of them came from foreign
data, and those were at room temperature, all very low
temperature.
So you've got to go back there and that was the
whole point. Besides those laboratory tests, those u-bands
or the samples that they had at MB-2, are the tube data that
was not run under typical conditions. Maybe some of them
were, but most were not.
DR. SHACK: I guess we could have a debate on just
how well you could make those corrections, but onward.
DR. CATTON: These things are scalable from one
pressure temperature to another.
DR. SHACK: I would argue that if you did it, you
know, what you do -- the thing that's undetermined in this
test is the crack area, the effective area. You can
determine that with one test under one set of conditions,
and then use the code to essentially extrapolate to other
conditions.
L over H is just -- you know, I have assumed the
simplest crack model. It's L in this case is a 50 mil wall,
and H is the crack opening. I should mention also if you
look at Amos and Schrock, his hydraulic diameter is 2H and
he can't divide L over 2H correctly, but we'll assume he
gets the thermal hydraulics right.
DR. KRESS: Would you repeat that?
[Laughter.]
DR. CATTON: I want that circled in the record.
DR. SHACK: Let's talk about circumferential
cracks. Again, the presence of a crack in a pressurized
tube produces bending, and the behavior can depend strongly
on whether this bending is constrained and on the fracture
toughness of the material.
And you end up with a fairly complicated looking
plot that looks something like this. And, again, if you
take a single crack or a tube with a single crack and you
pressurize it, what will happen is, it will bend.
And so the failure for that is this so-called
free-bending solution that you see right along here.
Now, the other thing we want to note is that for a
pressurized tube, if you're less than 100 degrees, you've
got a crack or not, this thing is going to fail in the axial
way, simply because of the 2:1 pressure ratio you have in
the tube.
So, you know, circumferential cracks don't even
start to enter the picture here until you've got 100
degrees, and then it matters whether you've got the
free-bending solution or what I have called the
fully-constrained solution that is when you constrain it
against bending.
You could do that with a teflon line or in a tube.
The easiest way to do it, experimentally, is to put two
symmetric cracks, one on each side of the tube, and then
you'll essentially have a fully-constrained solution and it
will look like this.
And so you'll be able, at any particular load --
or you can have a much bigger crack before you get failure
if you've got the symmetric loading, than you do if you have
the free-bending case.
Well, in the steam generator, we don't have free
bending, and we don't have fully constrained conditions.
We've got a tube support plate, you know, some couple of
feet above this thing, and we've got a tube, so this tube
has some flexibility.
And this is all covered in this Parameter C that
we have here. This is sort of a stiffness measure, and it
measures how much restraint you have, that that's a function
of the stiffness of the tube and the length that you have
between the supports.
In fact, the condition, if you assume it's simply
supported at both ends, or you assume it's constrained and
built in at both ends, if you look at steam generators, you
will find that this value of C is really somewhere between
.3 and .5. That would be a typical value.
And so that means that your curve sort of looks
like this.
DR. KRESS: What are you plotting?
DR. SHACK: I am plotting the pressure versus the
crack angle. And I want to know at what pressure will this
crack begin to extend unstably to grow? So it says under
normal operating conditions, I can have a crack that's 340
degrees through-wall, before it begins to extend.
DR. BALLINGER: We have an emaciated version of
that figure in the handout, at least mine is.
DR. SHACK: Oh, how interesting.
DR. KRESS: That's why I was asking you.
DR. SHACK: That's what happens when you send
McIntosh figures to people printing them from Word and a PC.
CHAIRMAN POWERS: I'm going to have to confess
that even in the fully-developed McIntosh version of it, I'm
a bit lost on this figure.
DR. SHACK: Okay.
CHAIRMAN POWERS: You're not plotting pressure
against something; you're plotting something normalized.
DR. SHACK: Right, the pressure over the burst
pressure of the unflawed tube. And so think of a curve that
comes from about here down to here, and then it goes to
here.
That's the failure curve for a steam generator
tube with a circumferential crack.
CHAIRMAN POWERS: Okay, now, on these squares and
diamonds and circles, are those datapoints or simply
indicators of some calculation?
DR. SHACK: No, that's -- I will get my staff to
get out of the bad habit of putting symbols on curves that
are purely calculations.
DR. CATTON: Those are purely calculations?
DR. SHACK: Those are purely calculations. So
those are calculated curves for a range of stiffnesses.
This would correspond to the distance of the tube support
plate, to the -- from the tube sheet, and, again, as I say,
for a real steam generator, the number is about .3 to .5.
DR. CATTON: For a given steam generator tube, you
can calculate to C?
DR. SHACK: Yes. I didn't think you -- I can give
you the formula for C, but this is a viewgraph.
DR. CATTON: I understand.
CHAIRMAN POWERS: What is the sigma, sub-Y over
sigma-super-bar?
DR. SHACK: That's essentially the ratio of the
yield stress to the flow stress in this material, and we've
got a power exponent, so it's a power law hardening material
with a power hardening exponent of .18. We're allowing
plasticity in this solution.
DR. BALLINGER: This flow stress is done by the
yield plus ultimate over two?
DR. SHACK: Over two, right. Again, I can give
you a detailed reference for the solution for the
circumferential support.
CHAIRMAN POWERS: Is ultimate over 2, so the ratio
of the yield to that obscure thing is a half, which means
ultimate and yield are the same?
DR. SHACK: No, no.
DR. KRESS: That is a material property is what
you are saying?
MR. BALLINGER: No, that is a rubric. Yield plus
ultimate over 2 happens to work.
DR. SHACK: I will take it back, I am not sure --
DR. KRESS: It is just the average.
DR. SHACK: This is something describing the power
law hardening curve that was used for these calculations.
This is -- we have done this three ways, with an elastic,
perfectly -- or an elastic, rigid plastic material, an
elastic tangent modulous material and a power law curve.
Exactly what this power law is, I will have to back and
check.
MR. BALLINGER: The yield plus ultimate over 2 is
used in general in all these calculations. It just happens
to work.
DR. KRESS: It just happens to work.
MR. BALLINGER: It just happens to work.
DR. SHACK: And Westinghouse and I, we will fight
over whether it is .5, .55 or .595.
DR. KRESS: But this is because you are failing in
flow plasticity.
DR. SHACK: Plasticity.
MR. BALLINGER: Plastic, it is a fully plastic
case and so yield plus ultimate over 2 is about an average
value for the flow stress.
DR. KRESS: About an average between, yeah.
MR. BALLINGER: And it works for strain hardening
materials.
DR. KRESS: Right.
DR. SHACK: Yeah. I will have to go back and
check this.
DR. KRESS: See, you have to explain these things
to us thermal-hydraulicists.
DR. SHACK: But the important thing is that you
can have extraordinarily large circumferential cracks in
this material.
Now, let's go back to the main steamline break and
some of the additional loads where, you know, I am only
calculating the pressure loads here. You know, these tubes
are thin wall tubes, so any axial force that I put on it
doesn't produce any hoop stress. If I have an axial crack,
without any change in hoop stress, I am not going to -- I
can change the axial stress, I am not going to do anything
to open that crack or to fail cracks in that direction. I
mean that is one of the fundamental assumptions of linear
elastic fraction mechanics is that I can have Mode 1, Mode 2
and Mode 3 cracking and they are independent.
DR. KRESS: So I don't have to worry about thermal
stresses then?
DR. SHACK: You do if you have thermal stresses
that will give you hoop stresses, but if you have thermal
stresses for axial cracks, if I have axial loads,
essentially, they have no effect on the axial crack. Now,
that is not quite true, there is kind of a second order
effect in this curvature thing, if you notice that bulge.
If I put an axial tensile load on here, I actually restrain
this tendency bulge by kind of a cable sense. And if I put
a compressive force on here, I would make it go a little bit
more. So there is a second order effect in a circular tube
under bulging because of that load, but that is a second
order effect, you know, it is pretty small.
DR. KRESS: That is when it already starts to go,
as opposed to whether it will go at all.
DR. SHACK: Right. Well, it could even have a
small effect on whether it starts to go, but I mean you
would really have to believe your calculations out to more
significant figures than I believe these models to worry
about that.
So I would argue that for the axial cracks, the
additional loads I might get under the main steamline break
will have very little effect on the crack opening or any
potential failure of those axial cracks.
MR. BALLINGER: The only complication might be in
the U-bend.
DR. SHACK: The U-bend. Well, again, we are
talking here 95-05 considerations, where we are in a
different beast.
DR. CATTON: The vibration caused by the event,
that is going to rattle them in every way.
DR. SHACK: But it is not going to put in this
kind of mode, the bulging mode for a circular tube. I am
going to have all sorts of bending modes, but all bending in
long, thin wall tubes produces axial stresses, you know, and
that is not true if I bend it enough to make it into a
U-bend, you know, if I turned it into a pretzel. But, you
know, these have been designed for these loads, I am not
going to get that kind of plastic deformation. You know, I
don't expect the steam generator to come apart and the thing
to bend over in a 90 degree bend. But, otherwise, I am not
-- I don't get coupling between the axial and the hoop
modes.
So the axial cracks, I don't really expect any
real major effect of the additional loads that I get from
the main steamline break.
Circumferential cracks, well, in the 95-05
context, --
DR. CATTON: When you make these arguments, what
kind of loading do you have in mind taking place inside the
generator? I can envision --
DR. SHACK: I am assuming it is not large enough
to fail the tube intention, yes. I mean if I had loads big
enough to fail the tube intention, I don't care whether I
have an axial crack or not. And, again, you know, the
blowdown loads here are not -- I don't exactly know what
they produce.
I know these things were designed for them, and I
know the way the code designs it, so I am assuming it was
designed to have perhaps a limited amount of plastic
deformation. You know, they would have somewhat relaxed
design criteria. You know, it wouldn't be pressure vessel
stresses, but it would be limited to some level.
Now, again, you can't make quite the same argument
on the circumferential stresses because I have axial
stresses now, and they act on circumferential cracks. But
in the 95-05 context, again, you have some circumferential
cracking in the tube support plate, but it is really
predominantly axial cracking if you look at all the
metalography. Much of the circumferential cracking is this
so-called cellular cracking, which is a kind of cousin to
IGA. Much of it probably is fairly shallow, is not
throughwall.
So, again, you have got -- and, as I mentioned, if
I had the same size axial crack throughwall, and the same
size circumferential crack throughwall, it would take three
times the stress on the axial, to open up the
circumferential crack as much as it would the axial crack of
the same length.
DR. KRESS: Why is it?
DR. SHACK: Because one is in a curvature and one
is in a flat plate.
DR. KRESS: Oh, I understand that part. But why
is it you get more axial cracks, a lot more axial cracks
that you do circumferential?
DR. SHACK: Oh, because I have got a 2 to 1
pressure ratio in the tube. You know, in the tube support
plate especially, the stresses, unless you have big dents,
which is a separate problem, is really the 2 to 1 pressure
stress that I have.
I get most of my circumferential cracking in these
things at places like the roll transition, where I put in
residual stresses which can be just as large in the one
direction as they are in the other, but overall, I mean that
is why can tolerate these mother cracks. You know, there is
nothing, this material is non-isotropic. It is not stronger
in the axial direction than it is in the hoop direction.
You get a head start because I am only putting half as much
load on it in the one direction as I am in the other.
Now, the other thing that does come in is the fact
is that in this direction it is a flat plate, and in this
direction --
DR. KRESS: It is a curve.
DR. SHACK: It is curved.
DR. KRESS: That is what I thought you were
explaining.
DR. SHACK: Yeah, and you get both of those
working together to make a difference.
CHAIRMAN POWERS: Bill, I must be particularly
dense today, or maybe typically dense, but you come to a
conclusion down here at the bottom of this that says that,
gee, even under MSLB conditions, throughwall cracks remain
stable until greater than 300 degrees extent. Is that just
an assertion, or am I to derive this out of this figure?
DR. SHACK: Main steamline break hits the
instability line at 312 degrees.
CHAIRMAN POWERS: Okay. Now, I didn't understand
that that was an instability line.
DR. SHACK: Yeah. The instability line, again,
comes down line so. So if I had very, very high, high axial
-- and that is the other thing now here, again, --
DR. KRESS: Below that, you get the crack may go
through.
DR. SHACK: I am assuming this crack is
throughwall already.
DR. KRESS: Already.
DR. SHACK: And all I want to know, if it is going
to get longer.
DR. KRESS: You are just trying to make it bigger.
DR. SHACK: I am just trying to make it grow.
DR. KRESS: Okay. So, below that, it is stable at
the size it is. And above that, it is going to run.
DR. SHACK: Right. So, again, if I had a 200
degree crack, I can put an awful lot of extra load on this
thing. Again, I don't know how much I get in these things,
but I can put an awful lot of extra load. And I really
don't think that Gary and Jack are going to allow people to
operate with 200 degree cracks circumferentially.
CHAIRMAN POWERS: The problem is that their
detection ability of sort of circumferential cracks is much
lower.
DR. SHACK: But, again, in the 95-05 context, big
circumferential cracks are very, very unlikely and have
never been seen. You know, big circumferential cracks occur
at the tube support plate, I mean the tube sheet, the roll
transition.
DR. KRESS: Now, this whole discussion has to do
only with circumferential cracks, right?
DR. SHACK: Yeah, I did failure for the other
cracks back in this diagram.
DR. KRESS: That was the unstable.
DR. CATTON: We also had to put an adjunctive in
front of MSLB, "mild."
DR. KRESS: But on the other diagram, the previous
one, Bill, go back to the previous curve. I am doing it,
Dana.
CHAIRMAN POWERS: Well, that is good because I am
totally perplexed on these figures.
DR. KRESS: Where is your P for main steamline
break? Oh, you have got main steamline break calculated
separately, okay.
DR. SHACK: Right. Then I show the three curves
together here to show you the sort of different ranges of
crack geometries that are of interest if you are in the
operating range, in the main steamline break range, or the 3
delta P. So the 3 delta P requirement essentially removes
this range of cracks.
DR. KRESS: So the conclusion we draw from this
main steamline break figure is that you have to have pretty
deep cracks, like 75 percent throughwall, before a main
steamline break increases its flow area.
DR. SHACK: Right. Well, you have to have more
than -- you have to have 75 percent throughwall before the
crack will even pop throughwall.
DR. KRESS: Oh, yeah, that is right.
DR. SHACK: And, again, so if I had a long enough
crack at 75 percent, I would go through the wall, and I
would go -- but it so long, I don't care whether it is
unstable or not. You know, a leak that big is -- I am
already dead.
DR. KRESS: You are already dead.
DR. SHACK: Here, to get a leak from smaller
cracks of interest, I have to be .8 to .995 throughwall.
DR. KRESS: Which kind of tells you you don't need
to worry about main steamline break imposed loads for either
axial or circumferential.
DR. SHACK: No, no, that is not the message. The
message is that only certain cracks fail.
DR. KRESS: Oh, I see.
MR. BALLINGER: The fact is that you can miss a
long, 2-1/2 inch crack --
DR. KRESS: You can miss it, it might be there.
MR. BALLINGER: that is 70 percent throughwall.
DR. KRESS: And it is going to go through and
leak.
MR. BALLINGER: And then it will rupture.
DR. SHACK: Again, here is my implications from
all this again. I have left everybody confused, but here is
what I draw from this anyway. I am going to argue that,
again, talking more generally now, not in the 95-05 context,
that the primary mode of interest is this stress corrosion
crack. It is going to be associated with regions of high
residual stresses or aggressive chemistries.
The places that I am going to find that are the
tube support plate where I have crevice conditions that
promote aggressive chemistry. The roll transitions, again,
I have got high residual stresses there, I can get cracks on
the ID, I can get cracks on the OD, I can get axial and
circumferential cracks. Roll transition is a bad place.
Small radius U-bends, I get residual stresses
introduced during the fabrication process simply in bending
this thing around to make a U-bend, and as that radius gets
tighter, the stresses associated with that operation get
higher.
DR. KRESS: Are these steam generators small
radium U-bends?
DR. SHACK: Yeah, this is -- think Row 1, Row 2.
DR. KRESS: Row 1, all the ones right in.
DR. SHACK: Yeah, right. You know, are the tight
ones, I could have said it that way. You get additional
stresses if you have got hour-glassing of the flow slots by
denting and you move the legs of those things together.
And, again, I would argue that the cracks in the
small radius U-bends have the greatest potential for gross
failure. In the tube support plate, your cracks are limited
by the thickness of the tube support plate and opening and
leakage is constrained by the tube support plate, except
perhaps in main steamline breaks.
The high stress transition at the roll transition
is limited in extent, it is typically less than 10
millimeters. So I am going to get axial cracks that are
fairly limited in length, although I can get big
circumferential cracks, but I have argued that I can
tolerate pretty big circumferential cracks.
So, of the three main regions here, the small
radius U-bend, as Ron said, I can have a four to five inch
long crack, I have got a high stress region that is long in
the small radius U-bend, so I can get a big crack.
Now thoroughly confusing everybody, let's move on
to high temperatures, where I can really do it.
MR. STROSNIDER: Bill -- this is Jack Strosnider.
I was wondering if I could just interject a though before
you do move on to that.
I mentioned yesterday when we talked about the
steam line break issue that I didn't see this necessarily as
a Generic Letter 9505 issue. I thank Bill. I think he has
provided some quantitative arguments in that regard.
When I first looked at this issue, the thing that
comes to mind is exactly what Bill said. My concern would
be stress corrosion cracking at the top of the tube sheet in
the roll transition where we have had some significant
circumferential cracking.
The one thing I wanted to make you aware of is --
or a couple of things -- is inspections that licensees are
doing they are using rotating pancake coil probes at the top
of the tube sheet. If they know they have got that cracking
going on, they basically do 100 percent. In their initial
inspections by EPRI guidelines they would be doing 20
percent and if they find something they expand it to 100
percent.
The other thing I would mention with regard to the
fracture analysis here is Bill -- the analysis that is
presented here is dealing with planar cracks. Actually when
you look at these cracks that are occurring in the roll
transition they are not really planar. They tend in that
residual stress field of about a quarter to three-eighths of
an inch to be offset as you go around the tube and actually
some of the testing of those tubes in situ and where they
have been removed show that they have very, very high
failure strengths because of the ligaments that are there,
that they will leak, all right, and quantifying that leakage
is another question, but in terms of actually failing it
they do --
DR. SHACK: Just to expand that little bit, Joe
showed you a figure yesterday, it's in his presentation, of
a probably more realistic depiction of circumferential
cracking at a roll transition where he had four parallel
rows of cracks sort of spread out across the roll transition
and they went 360 degrees but they were segments, so -- and
as Jack said, when the guy does the normal kind of
inspection he is going to see that as a 360 degree crack.
It is going to look horrendous to him but when you see the
detailed resolution of that thing, it is really a whole
bunch of short little cracks and I suspect if we blew that
tube up we would find it probably had a pressure stress of
6,000 to 7,000 psi.
The other thing that we have seen --
CHAIRMAN POWERS: Well, let me interject here.
If you expect us to take this into account we're
going to have to see the data and if you are arguing for
taking a stand on high pressure we are going to have to see
that.
MR. BALLINGER: The actual field experience has
been that apart from fatigue failures there has not been a
tube rupture, correct me if I am wrong, due to a
circumferential crack other than fatigue.
CHAIRMAN POWERS: We have got 11 incidences of a
tube rupture. That does not constitute a database that
seems to preclude this.
MR. BALLINGER: I said field experience, not
database.
MR. STROSNIDER: I think we can provide some data
from the Maine Yankee experience I think where they did
some, my recollection is some in situ and maybe some pulled
tube tests and they did some metallography on this. We will
have to pull that out for you.
The final comment just for you to be aware of is
that with regard to circumferential cracks at the top of the
tube sheet and for cracking in the U-bend and some of these
areas we are talking about the plugging criteria is plug on
detection and anything that is detected is removed from
service.
Then you get back to what is the threshold of
detection and we had some discussions on that yesterday, all
right, and so anyway I just wanted to interject those
thoughts and we can provide some information on the cracking
at the top of the tube sheet.
CHAIRMAN POWERS: It seems to me that in our
discussion the probability of detection would -- I came away
with the impression that you difficulties in detection are
precisely in the areas that this slide says are our greatest
concern, the U-bend and the top of the tube sheets.
DR. SHACK: We have got the tube support plate
9505.
MR. STROSNIDER: And I would also point out that
with rotating pancake coil inspections at the top of the
tube sheet and the inspections that people are doing there,
and again recognizing the forgiving nature of those
particular defects I think we are in pretty good shape
there.
Clearly there are some issues in the U-bend. We
got Indian Point 2 in February where there was clearly a
threshold of detection problem. The crack that failed was
there in the last inspection but the quality of the data was
so noisy that they didn't pick it up, and that is something
we are dealing with.
The industry is currently working to incorporate
some noise criteria if you will into the EPRI guidelines and
we are working on a generic communication on that same
subject.
CHAIRMAN POWERS: And one can't help but wonder
how many more of these discoveries we have to make before we
come away with the enthusiasm that we should on this
detectability issue.
MR. STROSNIDER: Well, the only point I would
make, and I think Ken Karwoski -- you can paint a very dark
picture if you want, but I would also go back and look at
the actual data on the decrease in the number of leaking
tubes, the decrease in the number of tube failures.
If you look at those failures that we are talking
about, it is a large number up through 1993 and one since
then. It may not be statistically significant but I would
suggest that the advances that we have been talking about in
the inspection methods and the programs that are being
implemented are having an effect, so I wouldn't paint too
dark a picture.
CHAIRMAN POWERS: I would be interested in looking
at the number of tube rupture accidents that we have had on
a per year basis and see if that has come down equivalently.
MR. STROSNIDER: Say that again?
CHAIRMAN POWERS: The number of tube rupture
accidents that we have had --
MR. STROSNIDER: One for five years --
CHAIRMAN POWERS: Which is about the same rate
they have been going on before, so I mean nothing has
changed on that.
MR. STROSNIDER: It doesn't matter whether it is a
40 percent through-wall criteria or --
DR. SHACK: Well, as Jack pointed out, the
criterion here is not 40 percent through-wall. It is plug
on detection.
MR. STROSNIDER: Right, but the point I made is
that the data, I agree, may not be statistically significant
in terms of the change of the rate of tube ruptures but I
would suggest t hat if you look at the frequency of
ruptures, if you look at what happened in the '70s and '80s
and you look at what happened in the '90s and you add a
little bit of knowledge about the new inspection methods,
the use of the plus-point probe, the 100 percent
examinations, the scope of what is being done, all right, I
can't show it as statistically significant but I would not
want to discount it.
MR. CATTON: I think before you completely close
it out, we have got to find out what happens with GSI 188.
That is really where it's at.
For mild MSLBs you give a very convincing
presentation.
MR. MUSCARA: I want to go back to this issue on
the detection of circumferential cracks.
You mentioned previously now that we know that we
expect cracks at the top of the tube sheet we can at least
do inspections in those areas, not with bobbin coils but
with pancake coils.
As I mentioned yesterday we are doing quite a bit
of work to quantify inspections in that area also and what
we find is, yes, there's difficulty detecting small
circumferential cracks but the largest circumferential
cracks PODs do not do that -- it's fairly high -- and so if
we are talking about a 340 DB crack that you need to open
up, those are not missed. The smaller ones, yes.
DR. SHACK: I didn't mention it and I can't find
the transparency at the moment -- oh, here it is -- the
other thing you want to note is that the leak rates --
again, this notion that these big cracks -- these leak rates
are still fairly small through these cracks again out to
100, 150 degrees.
You are not getting a lot of leakage out of the
circumferential cracks.
CHAIRMAN POWERS: Bill, I guess I really am dumb
today. You've got a plot of a quantity that on the
appearance of it is nondimensional.
DR. SHACK: Yes.
CHAIRMAN POWERS: Okay.
DR. SHACK: It is the area over the flow area of
the tube.
DR. KRESS: There's sort of a leak rate.
DR. SHACK: Sort of a leak rate. Multiply by 600
gpm.
MR. BALLINGER: Is that where it is normalized to?
DR. SHACK: That is the one number that everybody
seems to be able to agree on is that if you have the tube
cut you will get 600 gpm.
We'll figure over CRACKFLOW and Henry versus time
relaxation but 600 gpm out of the end of the tube seems to
be a number we can all agree on.
CHAIRMAN POWERS: But if I do that, then I get
some reasonable numbers, don't I?
DR. SHACK: Yes. Those are big cracks though.
CHAIRMAN POWERS: I guess I don't understand why
it is small. I mean when I do the multiplication I don't
come up with a small number.
DR. SHACK: It is a small number --
CHAIRMAN POWERS: The flow relative to 600 I'll
agree but --
DR. SHACK: It is also small for a 150 degree
crack. That is a big crack.
Let's go on to high temperatures.
We are looking at the failure steam generator
tubes during a severe accident. We have got -- we have done
these tests. Essentially we wanted to bound the kind of
things that we're predicting -- I will learn to spell this
thermal hydraulic one of these days --
[Laughter.]
DR. SHACK: -- which sort of predict that we have
a range of something like 3 to 13 C per minute, kind of a
heatup rate. If these ramps are sufficiently rapid we could
depend only on the burst properties and they'll be history
dependent. We could use a flow stress model.
If they are sufficiently slow we have to take into
account the pressure and temperature history. We use a
creep rupture model.
The thing that we have noted, at normal operating
pressure we account for crack geometry through a stress
magnification factor, MP. We have an extensive database to
validate that at those temperatures, but we find from
analyses that if we take the kind of stress-strain curve
that we expect to get at 300 C and the kind of stress-strain
curve we expect to get at a much higher temperature, much
less strain hardening, we find that it doesn't make a whole
lot of difference in the MP that we calculate, so that MP is
really a measure more of geometry than material properties,
and we can use it in high temperature and at low
temperature, so that is an assertion.
We have to sort of demonstrate then that it works.
We are going to assume that the MP factors we
derive from low temperature tests are applicable and we are
going to determine failure by a creep time fraction model.
This is a sort of linear damage rule where we kind of scale
the rupture time according to the stress and temperature and
so if we run tests at one temperature and one stress and
then we are doing a variable stress history we can
essentially integrate that fraction of the damage that
occurs at that particular stress and temperature and just
sum it up until we to get to one for failure.
The stress that is active here is the actual
stress time, this multiplier MP that we have determined
comes from the flaw geometry.
What do we do for the validation tests? We did
isothermal constant pressure tests. We did some tests with
deep cracks to test how well the MP model was doing. We did
constant ramp rate tests where we just ramped up the
temperature with either a constant pressure temperature ramp
or an isothermal pressure ramp, so we did the ramp tests.
Then we did prototypical tests under varying --
some of them were more prototypical than others but they all
varied.
Here are some results comparing the results we get
from essentially the Creep Model and the Flow Stress Model.
We're looking at two different ramps here that
we've called the EPRI ram and the INEL ramp, and you may
remember those from the good old days, and Steve may bring
them up again, or he'd probably rather forget them all.
But they were --
DR. CATTON: This was an increase in temperature?
DR. SHACK: Yes, this is -- you know, was
essentially a projection of the temperature during the
station blackout accident.
And --
DR. CATTON: By EPRI and INEL?
DR. SHACK: By INEL.
DR. CATTON: They're probably both too low.
DR. SHACK: Well, the answer is, they are
different, but we managed to predict both ramps. We do the
constant pressure ramp, so you give us the ramp and we'll
predict the failure. That's the message.
Circumferential cracks, we don't quite as well,
but we do it enough.
CHAIRMAN POWERS: Let me see if I understand. The
symbols here are the datapoints and the line is the
prediction model?
DR. SHACK: No, the line is -- below the line,
you're --
CHAIRMAN POWERS: Okay, that's 100 percent
correlation?
DR. SHACK: That's the 100-percent correlation
line.
Okay, one of the other quantities of interest here
is the crack opening area at high temperatures, because
we're worried about leakage at high temperatures.
So we've calculated crack opening area under 300 C
conditions. There were a couple of questions that came up
here.
Is there any creep crack growth that occurs before
this crack goes unstable? That is, if we've got a crack
that's existing and we're now heating up the tube, can the
creep crack growth just make the crack get longer, so if we
start with a quarter inch crack at 300 C, by the time we get
to 700 C, will it be longer than a quarter inch or will it
just open up.
And, again, we're petty sure the crack opening
area is going to vary with time, and we want to be able to
predict that.
The analytical predictions are based on a an
analogy between a power law plasticity model and creep
behavior. What we do is, we take essentially the power law
plasticity model and we replace the strain by the strain
rate in the creep solution.
And we've got power law plasticity models for
center crack plates. The difference between the axial crack
and the circumferential crack is the fact that you get this
additional stress on the axial crack because of the bulging
factor.
So what we've -- we can't do tests on an axial
crack at high temperature without an infinite amount of
money. But we can pull on a tube pretty easily at high
temperature.
So what we've done is pulled on the tube at high
temperature, but we've said that the stress we should use is
M times the hoop stress.
So we've essentially done the axial loading with a
much higher stress to account for the fact that we haven't
got the curvature, so we've replaced the curvature with
essentially a higher stress to get an equivalent model.
DR. BALLINGER: Can I ask what COD is?
DR. SHACK: Crack opening displacement.
DR. CATTON: I should have known that one.
DR. SHACK: Okay, well, this sort of just says we
can't do these tests on the through-wall axial crack tubes,
and it's under internal pressure and it would take an
infinite gas supply system.
We thought about doing it on cracked plates, but
then we decided that the easiest thing to do was to take our
tubes and just put some symmetrical notches on both sides.
As I mentioned, that puts them -- it's like a flat
plate, but it just happens to be a repeating flat plate with
a period of pie-D.
So, there is it. You've got symmetric cracks.
And that's basically equivalent to this flat plate solution
with two cracks.
And we're good, we can do flat plate solutions,
and we like those.
Then we did a couple of different kinds of tests.
We did these isothermal validation tests, where we just
heated the temperature up to near 700-C.
We put a load on it, and we predicted how the
crack would open as a function of time.
And so again we've got a constant load, we've got
a constant temperature, and the crack is just opening up as
time goes on. And so you can sort of see how it's going up,
and we've got the observations versus the predicted.
And we've done this at two different load levels.
CHAIRMAN POWERS: For higher loads, you started
with deviation? Is there any significance to that? I mean,
if I went to 3,000 pounds, would I see a much bigger
deviation?
DR. SHACK: I don't know. We would have to run
the test.
CHAIRMAN POWERS: You don't have an explanation?
DR. SHACK: I don't have a good explanation for it
now. Those we did with two 45-degree notches. We wanted to
go back and do some more sort of notches that we thought
would be more protatypical, which is a .25 inch kind of
thing, the kind of small notch that Steve worries about
opening up and losing flow out of in the high temperatures.
And, again, this is another one of these
isothermal validation tests, and, again, the way we do these
essentially, it's sitting in the furnace. We open the
furnace up, we peek in with the telescope, make the
measurements, close the furnace back up. We're doing the
cheap.
CHAIRMAN POWERS: The High Temperature Committee
developed better ways to do that, by the way?
DR. SHACK: You know, on our budgets --
DR. CATTON: Sometimes that's where the best work
is done.
DR. SHACK: Now, we wanted to do a non-isothermal
validation test, and in this case, we used the temperature
ram simulating six RU. This is probably the temperature
ramp that Steve will show you today. This is the one they
believe is the -- Joe will show you.
Now, of course, when we're doing the isothermal or
the transient, we can't open the furnace up. So here we
only get one datapoint.
You know, you hit it, baby, or you miss it, so
here's the temperature, here's the predicted notch
displacement as a function of temperature, but the only
point we can validate is the one right there at the end.
And, again, we were doing pretty well on the --
DR. BALLINGER: You're going to get -- an LA-600
is going to be well behaved in that respect, because you've
got that cliff at about 650 C where the yield strength drops
off like a stone.
So you're into the creep regime and it works.
DR. SHACK: Well, the other nice thing about this
that we're always surprised about is, in the creep regime, a
lot of this heat-to-heat variation goes away.
You know, that all arises from the differences in
the working that you've done, and you heat that up, and that
all goes away and we're sort of left with the basic,
fundamental crystal structure of Alloy 600, and so you get
much less material-to-material variation in the creep
regime.
If we just look at these things, they open up into
rectangles. You know, there's no creep crack growth here.
They don't get any longer, the suckers just move apart, and
they turn into rectangles.
So they started out as narrow slots and they
opened up as wide slots.
DR. BALLINGER: It's tough stuff.
DR. SHACK: Tough stuff.
Now, this crack opening area begins to, again,
increase rapidly. If you look at this crack opening area,
it sort of goes along, along, and as Ron mentioned, you
know, you kind of fall off this cliff around 650, and the
action starts to take place, so that basically there's not a
whole lot of increase in the crack opening area till you get
out to about 650, and then it starts to take off.
And so what we've done here is looked at --
suppose we had a final temperature of 700 C before something
else failed or if we had a final temperature of 750 C, you
can predict the crack opening areas, at the crack length at
those two temperatures.
You can also predict the leak rate through those
crack opening sizes, again, as a function of crack size at
the two temperatures.
CHAIRMAN POWERS: It's easy to compare these
because these are in kilograms per second, as opposed to
gallons per minute, right?
DR. SHACK: Well, I had them as gallons per minute
when I started out, but they told me that when we deal with
gases, we do it in kilograms per second.
Well, that was all I wanted to say -- well, let me
just -- we've got a couple of extra ones.
Life gets harder when you get to the real world,
of course, because when somebody hands me a real crack, it
never looks like a rectangle, unfortunately, and so you have
to make some sort of judgement as to how you're going to
model this crack in terms of an equivalent rectangle.
And there is a discussion of how to do this, and
there are some various procedures that we're trying, that
people use, and we've --
Without going through them, we're trying to
validate those kinds of procedures by looking at, again,
controlled shapes. It's easy to triangles and trapezoids,
and you kind of compare what you get in burst pressure from
the triangular and trapezoidal notches with essentially the
equivalent area kind of models that we're working through.
Again, that's more of a detail, I think, than we
need to get into here, but it is a question that has to be
addressed. And that's where we're going.
Any questions?
[No response.]
CHAIRMAN POWERS: Any other questions for Dr.
Shack? Anyone that thoroughly understands everything that
he's told us?
Okay, that's good. Thank you, Bill.
Are we -- did we exhaust the subject of crack
unplugging yesterday?
MR. STROSNIDER: I don't think we have anything
else to present in that area.
CHAIRMAN POWERS: Okay. It's just listed on my
agenda here, and I know we talked about it a lot.
One of the issues --
DR. SHACK: My L over H curve is sort of a crack
unplugging model. It's easy to unplug cracks of L over Hs
and too big. The bigger it gets, the harder I would suspect
it would be to unplug the crack.
CHAIRMAN POWERS: I don't pretend to understand
that.
One of the issues that falls under the general
nature of crack unplugging is probably also material coming
out of the crevice regions.
Do you have anything that you'd like to talk about
on that aspect of crack unplugging? That was an area that
we didn't explore yesterday.
MR. STROSNIDER: Are you talking about loss of
material between the tubes, the plate and the tubes?
CHAIRMAN POWERS: Right.
MR. STROSNIDER: Okay, I don't know if there is
anything for me to talk about.
MR. KARKUOSKI: Just in that area, the only thing
I would add is that when we do these leak tests, these leak
tests are performed as if that degradation is in the free
span so if there's any material that stays around the tube,
it would only serve to restrict the leakage. The
correlations are all based on free span tests.
CHAIRMAN POWERS: Okay. That's actually very
useful.
Okay, if that exhausts that discussion, then I
think we can afford to take a break till quarter after the
hour.
[Recess.]
CHAIRMAN POWERS: Let's go back into session.
We are now going to discuss the accident framework
for a lot of these technical issues that we have been
covering. We have made a distinction, appropriately I
think, between design basis accidents and severe accidents,
but to my mind some of these things cloud the definitions of
the distinctions that one likes to draw between design basis
accidents and severe accidents.
In particular, the essence of my challenges here,
it seems to me is in the design basis analysis one analyzes
a main steam line break and one analyzes steam generator
tube rupture accidents, and one is supposed to have a plant
that accommodates both of these.
Now one has a situation where a main steam line
break involves a steam generator tube rupture, which up till
now has never been done as a design basis accident, and so
design-basedness becomes a little more complicated.
One of the areas that it becomes very complicated
in thinking about actually comes back to the iodine spiking
issue, that in the past we have said okay, let's calculate a
spiking value looking at what the steady state coolant
concentration is according to the tech specs.
Now you have plants operating much lower than the
tech spec limits though they may have not changed their tech
spec limits. Even if they did change them, they are still
operating a couple more at the risk magnitude below.
Now if one hypothesizes that the spiking factor
that one has is inversely correlated with that coolant
concentration, it seems that if one follows the prescription
of design basedness, that's fine, but I still use the tech
spec limits following that, but that is a lower spiking
factor than one would have if one used the operational ones,
so things get very confused between realistic and design
basis.
Anybody that can help me understand these a little
better I would appreciate it. That is your cue, Gary.
MR. HOLAHAN: This is Gary Holahan. In fact, the
Staff will make a presentation on both design basis and
severe accident issues. Steve Long is going to start off in
fact trying to define what we mean by design basis
accidents.
I would say something a little different from the
way you introduced it, Dr. Powers, and that is I think that
we are still preserving the concept of design basis, meaning
looking at spontaneous tube ruptures and looking at steam
line breaks, not steam line breaks with tube ruptures, but
we are looking at steam line breaks with increased leakages
that may be associated or expected to occur given the main
steam line break.
We also have severe accident analysis which looks
at main steam line breaks and a whole spectrum of other
possibilities, some of which are quite unlikely but much
more serious than steam generator tube leakage.
I think we will cover both main steam line break
with leakage and main steam line break with tube ruptures,
in fact, multiple tube ruptures will cover all those cases,
but the more extreme cases we'll discuss this afternoon.
The other issue that I would like to make sure the
committee understands is on the viewgraph it said that Dr.
Parry would be here this morning, but in fact he will be
here this afternoon to talk about operator action and human
reliability analysis in the context of the severe accident
issues and if we have design basis human reliability
questions, which I think are very limited conceptually, I'll
either try to cover them this morning or relate those to
this afternoon's discussions.
CHAIRMAN POWERS: I think the issues of human
actions during design basis accidents are raised by the
statement of considerations where there is a phrase in
considerations that I am sure I can't quote accurately from
memory but it is to the effect that provided several key
operator actions are carried out, and I think that is mostly
controlling the usage of water during the accident, and what
happens it seems to me is the time available for making
those key operator actions can shrink under some of the
higher leakage assumptions associated with main steam line
break, so I think it's just a matter of understanding how
one decides that one can credit operator actions in light of
the time available.
That has been an area of some contention for some
period of time.
MR. HOLAHAN: The distinction that I would like to
make is in the design basis accident context those operator
actions are targeted to keeping the event within the dose
guidelines of Part 100 and so forth. Tube ruptures -- that
means isolating the leak. For steam line break I guess it
relates to the cooldown.
In the severe accident context the operator
actions are preventing core damage and so there are a
different set of considerations. As we go along we may pick
those out, but when they look like core damage issues I am
going to push them off until this afternoon.
DR. BONACA: One note however. Although the
design basis has the objectives you stated, the ERGs, which
are the emergency procedures that currently the Westinghouse
operators follow has consideration of steam line break with
consequential failures of tubes or depressurization of the
secondary side too.
I think it is important that in that context if
there is an opportunity we discuss those kinds of procedures
because clearly the operators are being trained for
scenarios which are not part of the design basis strictly or
the severe accidents. They are trained for intermediate
situations where in fact you have to bring the system down
to RHR and they are being trained to do that both looking
for a subcooled condition to enter the RHR or even in a
saturated condition, which means or implies a very large
break and opening to the secondary side.
I think at some point, and I don't know if we have
any expertise on the ERGs, but that would be valuable for us
to understand how they support the human reliability
analysis that is presented in the NUREGs.
MR. LONG: I'd just thank the whole group for
presenting the first slide. My name is Steve Long.
[Laughter.]
MR. LONG: We want to rearrange the order a little
bit here. Joe Donoghue would be up next to talk about the
ability of thermal hydraulic codes and then I would be up to
talk about the equilibrium between ECCS flow and leak flow,
and then we have deferred the next issue, on operator
actions, to this afternoon, and then Joe Donoghue would be
up again.
We have decided to simply this process.
I will go through the description of the
relationship between the flow from the ECCS system and the
flow out the leaks and then we will let Joe do the rest of
the subjects this morning.
The first thing, I think it is important to
understand the intent of the review that we did in NUREG
1477.
MR. HIGGINS: Steve, before you get off into that
detail, I had just one additional clarification on design
basis versus the other things.
Yesterday we talked a little bit about whether or
not this Generic Letter 9505 with the alternate repair
criteria and the analyses associated with that really
constituted a new design basis accident, new design basis
event.
I guess I am still not clear whether you consider
that that is or not or you are just changing the analysis
method but you don't really call that a new and different
design basis accident?
MR. HOLAHAN: I would call it the same design
basis accident with -- the only thing that is substantially
different is the main steam line break, instead of having a
1 GPM leak now has leakage based on the likelihood of a
number of cracks opening, so I would say it is the same
design basis event with a different set of assumptions -- so
it is main steam line break with leakage and a calculation
done to show that it means the Part 100 guidelines.
That analysis is part of the design basis. It is
part of the licensing basis, because, you know, a license
amendment ends up described in the FSAR just like the
original 1 GPM case.
MR. HIGGINS: Thank you.
MR. LONG: To try to draw that out a little bit
further, first of all, this was done before risk informed
regulation.
The intent is to not apply this type of permission
to leave a particular type of flaws in service to anything
other than what we expect is going to be a confined area of
the tubes within drilled hole tube support plates.
It doesn't apply to egg crates. It doesn't apply
to free span.
There's a problem with analyzing exactly how the
tube support plate would behave during a main steam line
break, so on the one hand, there is an effort to act as if
the tube support plate were to completely move off the
flawed portion of the tube and Generic Letter 9505 requires
that the probability of those flaw rupturing be small and
that the amount of leakage that would come out of those
flaws be such that you could still meet the Part 100 part of
the regulation.
So, as Gary said, we are not supposed to have a
steam generator tube rupture as a result of a main steam
line break, and the specification there was that the
probability not be greater than, the conditional probability
not be greater than .01.
Is that a new accident or is it a specification of
how improbable it has to be that there is a new accident?
You can, I guess, take your pick on your
interpretation of that, but I think the intent there was
really to try to keep within the guidance that we had for
having a low probability of failure under design basis
accidents and leakage that was within the guidelines for
design basis accident for dose from design basis accidents.
On the other hand, there was at the same time a
feeling that you would most probably have the tube support
plates actually confined to those portions of the tubes that
were degraded, so when we looked at it from a risk
standpoint we did not see a high probability of something
that would move the plates off, so at the time we did NUREG
1477 we really didn't have a risk assessment in 1477. The
risk assessment is counting on the plates remaining in a
position that confines the crack sufficiently.
That has been a difficulty for us in dealing with
the industry because the industry is frequently saying,
well, we analyze these cracks as if they were in the free
span; why can't we have permission to have them in the free
span? That brings up a bunch of issues that we really
didn't deal with because we were relying on them not being
in the free span, and we will get into some of those issues
this afternoon.
It gets a little difficult when we use shorthand
in terms of whether or not something is a design basis
accident or a severe accident. There's different uses of
those words and different groups of jargon and it is often
allowing you to make an erroneous leap into something not
intended, and we will just have to keep reeling those in if
they get made throughout the rest of the conversations.
Are we ready to go for the next slide, next
subject?
The committee asked for a justification of the
assumption that the maximum leakage rate would reach an
equilibrium with the injection flow during the main steam
line break that induced tube leakage.
The explanation for this has to go back to the
context in which the assumption was made. The original DPV
document indicated that there might be a problem with not
being able to detect flaws that were more than 40 percent
through-wall and therefore it requested that licensees
either abide by the 40 percent through-wall criteria or in
some way demonstrate that they could meet a main steam line
break with 80 percent of the tubes ruptured.
That was dealt with by the Office of Research for
awhile, trying to figure out how many flaws might go
undetected in the free span and how many of them might leak,
how much they might leak, and some efforts were made based
on some assumptions about flaw growth rate to determine what
amount of leakage rate could exist under these
circumstances, and the numbers were quite high. They went
up around 10,000 GPM for a large number of flaws with large
growth rates.
That was the point at which we picked this up.
We were trying to put it into a context where we
could start thinking about the risk.
The difficulty was trying to figure out how you
would get that much of a flow rate, because if you can
somehow break the flaws that much you are well down into the
LPSI injection path that is essentially a large LOCA outside
containment.
Without going into human errors or human success
probability and dealing with large LOCAs outside
containment, I just want to go to the justification of the
assumption that you asked about.
And it basically goes to the thermal hydraulics of
a main steam line break, and I'm just going to put up a
sketch because the things that are available as graphs
didn't show very well, and I hope this shows.
Okay, if you look at what happens as a function of
time to the pressure in the RCS and the pressure in the
steam generator, when you break open the steam generator,
the fluid in the steam generator is at saturation, so it
doesn't just drop as if it's sub-cooled with a little bit of
cover gas.
It evaporates; it boils, so it holds pressure up
until it cools itself by boiling and it depletes.
And that cooling brings down the RCS pressure
along with it, so that the differential pressure in this
part really stays approximately the same until you've really
stopped the cooldown process.
At that point, you've tripped your reactor coolant
pumps, but you still have decay heat. But the major
repressurization process is that you're pumping in emergency
core cooling water, and you may be trying to turn on heaters
in the pressurizer when you get level back to where you can
heat something.
So, at some point, you start getting a higher
delta-P. And it progresses in a reasonably quick manner,
but not an instantaneous manner, to a higher delta-P.
The question was, what would happen in the cracks
under this kind of a scenario? And neglecting the idea that
there are cracks and they might open for a minute, just
thinking about if there was a hole that suddenly appeared at
this point, it's very similar to a LOCA in the sense that
you're trying to pump water in and the leak is removing
water.
So you have a curve for the leak rate that's a
function of the pressure that's driving water out of the
leak, and you also have a function for the amount of water
that can be put in by the centrifugal ECCS pumps.
So, at low pressure, the leak is not going to be
putting out much and the pump is quite capable of pumping in
a lot of water, and as the pressure goes up, the pump is
going to be going to less and less input, and the leak is
going to have more and more driving force at the fixed area
hole.
So typically for LOCA analysis, you get to
whatever this pressure is, and it equilibrates there, at
least temporarily until you change something else in the
RCS.
So that's the kind of thought process that I want
to go to, but then I want to add the idea that the cracks
start with a very small hole and are increasing that hole
size.
So this is now not this curve, but as you increase
the pressure, you may be doing something like this as you
make the hole larger, as well as make the pressure greater
for driving fluid through those holes.
If you're starting off at a delta-P that's very
similar to what has been experienced for a long period
during operations, you know that the holes aren't opening up
very rapidly there.
However, the tests that have been done at the
National Labs, where they have taken cracked tubes, put them
into a test apparatus and stepped up the pressure, and had
hold times in the pressure inside the tube, have shown cases
where the tube may sit at a constant pressure without
leaking, and then suddenly without increasing the pressure,
it will start to leak, something will actually let go and
the leak will occur.
Or you may find that something that is already
leaking and is being held at constant pressure, will slowly
increase leak rate or maybe it will make steps in leak rate.
Now, we've seen all of these things occur. These
are happening, though, at small leak rates, and the leak
rates are staying small for one particular crack. It's not
a rupture, it's just a change in the crack opening area that
may not be a single value as a function of pressure.
And I think this is one of the major points in the
DPO.
And we tried to put that in the context of the
scenario where the delta-P in the reactor coolant system is
increasing, and we are envisioning a very large number of
cracks.
What we were envisioning was that these cracks
would not all behave in unison, so that if one of them would
pop, every one of them would pop at exactly the same moment.
And the wording is down here in the slides, but
rather than put it up and read it to you, let me just try to
talk my way through it.
The picture I was trying to come up with is
something that would tell me how far I could expect to open
the cracks before I'd really lose the driving force for
opening them any more.
And the logic was this: That if your delta-P is
going up in time from a value where essentially there
weren't any cracks open, and cracks begin to open, that the
delta-P is going up because you're putting water into the
system.
And as you open more cracks, you're removing more
water, and eventually you should reach an equilibrium
similar to what's going on here, but not necessarily at the
original hole size. You're increasing this curve just more
rapidly.
So you'd eventually come to some equilibrium point
higher than your normal delta-P across the steam generator
tubes, where you're putting water out at the same rate that
the ECCS pumps can put water in.
Okay, still, that's a constant pressure
differential higher then they have been experiencing before.
Maybe they can continue to pop and tear open a little bit
further.
So for that process, again, if you increase the
area more, you raise the curve up, so it's now running up
here, rather than down there.
The pressure drops, more water comes in from the
ECCS system, and you can envision that perhaps reaching a
point where so many things have opened up that you've gotten
all the way back down to the normal delta-P, in other words,
you now have essentially zero pressure on the secondary
side, your reactor coolant system is now at a pressure that
was equivalent to the pressure difference between the steam
generator and the RCS previously.
At that point, we didn't see any reason to open up
the cracks any further. They were stable at that point.
Usually in the laboratory, if you've pumped a
crack to the point where it starts to leak, and you drop the
pressure substantially, the crack pretty well stabilizes; it
doesn't continue to come apart, unless you've lowered the
peak pressure differential that it's in.
So the argument is essentially that we don't see
any mechanism for the delta-P in the system to open cracks
beyond the point that the ECCS pump could support when the
ECCS pump back pressure is equal to the original steam
generator delta-P.
CHAIRMAN POWERS: But this is a conclusion one
raises because you're looking at a very quiescent system?
MR. LONG: There's --
CHAIRMAN POWERS: When we look at a system that's
producing sonic booms and pressure pulses and things like
that, maybe those arguments aren't so strongly made.
MR. LONG: Okay, well, first of all, the sonic
booms and so on should be stopping in time, somewhere down
in here. So in terms of timing, we're not expecting that to
necessarily be concurrent with what I was just talking
about.
So the limitation on this is, if during this part
of the process here, you're talking about the new generic
issue designation that cracks that have been initiated as
stress corrosion cracks, are now being fatigued by
vibration, that's a different phenomenon.
And the thing you've asked me to justify was not
intended to try to cover that kind of phenomenon.
Now, earlier on in the process, when we were doing
NUREG 1477, I was talking to Joe Hopenfeld about this and
some other things, and he was discussing vibration as being
one thing that would open them.
At least insofar as I was hearing it, I was
hearing it as vibration being able to shake the plugs out of
cracks that were plugged with crud or something of that
sort, as opposed to the fatiguing issue.
So to some -- let me just finish the sentence. To
some degree, if you're dealing with cracks that are not
being increased in size, you can go up into this section of
the curve and say, well, if I'm starting to leak here, then
what I'm going to do is drop this pressure even faster on
the RCS.
And I'll be dropping my strain again. The
difficulty we had was we didn't have a mechanism that we
could use to show us how much we could open these tubes,
other than the strain from the pressure.
We talked about things, and I think a lot of
people in this room had to put up with me asking them
questions about if we had a large number of cert cracks and
there was a displacement by the upward force, can you
essentially pull apart a large number of cert cracks?
We were looking for things that weren't
self-limiting in some way, but we didn't find something that
we could physically credit and put a conditional probability
to and put into a risk assessment.
So, essentially, this is the description of what
we were thinking of the time, and what we think it was good
for and what we think it wasn't good for.
I'll answer questions on that and turn it over to
Joe Donoghue.
DR. HOPENFELD: I have a minor comment. In that
original document, there was a description of that droplet
eating the adjacent tubes, if you remember.
MR. LONG: That's also true, and we at that point
were not thinking about the droplets eroding the tubes under
main steam line conditions.
We were worried about it under hotter temperature
conditions. And so we weren't crediting that one, either,
for this particular analysis.
MR. STROSNIDER: Steve, this is Jack Strosnider.
For the system response that you're talking about
here, does it really matter how the leakage -- where it
comes from?
I mean, you were talking about assuming that
there's a hole and that there is some leakage, right. And
this idea of the system equilibrating at some point, does it
matter if it comes from 9505 leaks or if it comes from
hyperation or anything else, right?
I think you were trying to address the issue more
of a system response to a leak; that's the point.
MR. LONG: Well, there is a difference. If the
only thing that's creating the additional leakage is the
additional delta-P, then what the tubes have demonstrated is
an ability to survive that for a long period of time.
If there's no other driving mechanism besides that
elevated delta-P, you can make this limitation and say if
it's your ECCS pumps that are providing that delta-P, you
can follow your pump curve and figure out how much your flow
rate is going to be at worst, that you would have to deal
with, and how fast that will deplete the RWST and so on.
If you have something that's mechanically damaging
the tubes, even though it requires a delta-P to do it, it's
a new damage mechanism, you know, some sort of additional
tension or vibration or whatever that might, along with a
delta-P, create more damage to the tubes than they have been
seeing when they were in a quiescent condition at a delta-P.
You might break them open further, and you might
get more flow rate. You might go farther, but --
DR. CATTON: But that poor flow rate is still
going to be a function only of a delta-P. That's just that
it now is the square root of delta-P, and now it's going to
become maybe proportional to delta-P, because two things are
happening: The area is getting bigger, so it's just a more
complicated control valve; isn't it?
MR. LONG: It's more complicated than I predict a
limit on, is my point.
DR. CATTON: Still, if the pressure drops back
down, it's going to shut back down; it's going to slow down.
MR. LONG: If what you're doing is -- I'm
speculating here. If --
DR. CATTON: Well, I was, too.
MR. LONG: If you have fatigue cracks -- if you
have cracks that are growing by fatigue, you know, from the
vibration, and the fact that they're pressurized internally,
what's the limit on how much you can open cracks in the
system? How much delta-P do you need to keep opening the
cracks?
Just because you've gotten down to the delta-P
that they were stable at before you had the vibration,
doesn't mean that the with the vibration continuing, they
would remain stable under that condition.
DR. CATTON: So when you look at this curve,
wouldn't that just mean that you wind up staying down at the
bottom?
MR. LONG: You're saying that this winds up down
here?
DR. CATTON: The pressure doesn't go back up.
MR. LONG: Okay, but if you found that you've
leaked and then you go back up, then what you're really
saying is -- and there's still a delta-P along here that's
--
DR. CATTON: You empty your IWRST and you're in
trouble.
MR. LONG: Well, that's part of it, but I don't
think you can say you'd stay there.
If you had the same delta-P that you started with
or just a little bit more, and you're shaking the tubes.
DR. CATTON: The more open space you've got
between the two systems, the smaller that delta-P is going
to get.
MR. LONG: Right, so what you're really saying is
not that I get here, but that this comes down here.
DR. CATTON: That's right.
MR. LONG: That's my point. I don't know how low
to say this would go.
DR. CATTON: It depends on how big the area is.
If you make it big enough it will go all the way.
MR. LONG: It depends on how much damage you get
from the vibration. So my limitation on this is that I
can't say that the flow from ECCS pumps at a particular
value, which is this delta-P, is the maximum flow I expect
from the primary to the secondary.
I've got to take into account, the damage in some
other way, if that is the damage mechanism.
DR. BONACA: I have a question that I would like
to ask: It seems to me that we can argue about the damage
mechanism forever, because there is a position that says we
are going to have as much damage as you want and as much
leakage as you really can postulate on many tubes.
And there is a position they are presenting where
the leak is self-containing, and this must be a leak on the
order of one tube, maybe two tubes, because you're showing
pressure coming back up.
And if you had much substantial more failure
there, pressure would not come up, back again. I mean, it
would stabilize somewhere pretty low.
It seems to me that if we are trying to ask the
question, will the operator be able to deal with the
leakage, whatever leakage will come out, and what kind of
range for this kind of damage, it's such that the RWST will
not be emptied, and we will not come to a containment bypass
situation.
That's a central question, it seems to me, and so
the issue is have we looked at other flow rates that would
result from larger breaks or a larger number of breaks?
I mean, I have been reading a lot of these
reports, particularly NUREG 1477, and the INEL report, and
they seem to present a model where they have looked at up to
20 tubes failing.
So I would like to hear about that. I mean, if we
concentrate on the issue of will it happen or not happen,
we're going to be left with the dispute in place.
MR. LONG: One of the interesting things that
happens to me as I try to put all these things together into
a risk assessment is every time we run into one difficult
question, there is always the urge to bypass that question
by going to another area of study, and that turns out to
have a difficult question as well.
So if we assume that the flow rate will go to a
very large value, the other limit on that value that you can
postulate is essentially the size of a hole that is in the
main steamline as a flow restrictor, and that is a pipe
that, depending on the size of the plant, I understand is
from like nine to 16 inches in diameter, that would
basically run through the containment wall from a point that
is high up in the RCS.
We have thermal-hydraulic analyses that would
indicate how the plant system would behave under those
conditions and, essentially, it depressurizes quite rapidly.
It gets quite cool because you are doing a wonderful job of
cooling, and you are depleting the RWST very quickly.
If you look at the conditions, you are at RHR
entry conditions when you have a very massive leak like that
rupture. The question then becomes one of human error
probability, or even feasibility, if you look at the
guidelines. Can you actually turn on RHR under those
conditions? And it doesn't become just a matter of looking
at the procedures and the time available, you have to start
asking questions at this point about, well, where did all
that water go? If you have just emptied pretty much a steam
generator and the reactor, and two-thirds of your RWST out
into the plant somewhere, can you go turn on RHR? It
usually requires you to do something outside the control
room.
MR. HIGGINS: But, Steve, if we get into these
discussions now, haven't we left design basis accident space
and entered severe accident space?
MR. LONG: Yes. So, well, that is the --
MR. HIGGINS: And I didn't know if you had
transitioned in your presentation yet.
MR. LONG: That was the point, we are going to be
talking on the hairy edge the whole time for the rest of the
day, and I don't think we can just keep saying, well, that
is a severe accident, we will talk about it later. We have
to talk about the transition.
DR. BONACA: The reason why I asked you the
question, however, wasn't that I say that you have to
assume. We are trying to understand what are the
limitations of the combined power plant systems and operator
that will probably give us success up to a certain break
size. And then we will judge as reasonable people how
credible that size of rupture is going to be, and if it
bounds the concerns that have been expressed about damage,
or if it doesn't bound.
And, for example, one could say that if you
postulate a failure of 10 or 15 tubes, and you could make a
case where you can still give some success to the operator
in preventing the bypass, it would be more comforting than
saying that the operator cannot cope even with two tubes.
Okay. So I would like to just simply see if we can, at some
point, understand that, because I think that is an important
issue, and it tells us what we are dealing with insofar as
uncertainty.
MR. LONG: I think you are correct in wanting to
look at the human error probability part of this, and when
we get into the discussion later this afternoon, I will show
that I think that is an important aspect. I think it is a
dominant aspect for a lot of these things, not just the one
you are talking about now.
However, trying to use human error probability
calculations to narrow your focus for thermal-hydraulic
calculations doesn't -- usually it works the other way
around because we are a little more precise with the
thermal-hydraulics than they are with the human error
probabilities.
But it is a problem of can you get information
together to bound the issue or not, and it is a struggle
here. And one of the things I think you have to go back to
is, is there a credible method for making a large hole,
rather than just assuming a large hole? In the design
basis, we have chosen to assume large holes and required
licensees to do fairly significant demonstrations that they
can cope with those large holes in those places. But one of
the ones we never did require them to cope with is a large
hole that takes the RCS fluid to somewhere where it cannot
go into the recirc path, and that is something we have known
as a concern since the reactor safety study in the '70s.
I think one of the things you have to do in trying
to bound this whole question is approach all the pieces, not
just leave one go and try to do it with a few that remain.
And I think you have to look at, what do we think we can
really expect to get in the way of a hole size? What is
credible? Because if it is really a credible hole that we
haven't considered before, maybe we need to change the
design basis to include it.
On the other hand, when you get into risk, if you
think it is plausible, but not really high probability, you
may be able to handle, put in some of the rest of the
features and decide that the risk is low enough overall to
not have to go any further in the analysis. A risk model
does not define all things to a fine degree, a risk model
usually goes as far as you need to go to make a decision and
stops, hopefully, just a little bit beyond there, as opposed
to just short of there, to support the decision. And it is
hard enough to get to that point.
MR. HOLAHAN: Let me come back a bit. We are
talking about design basis, we are not speculating about
some, you know, future design basis. What we are talking
about is design basis, you know, as it is allowed in 95-05
or other situations, and none of these cases allow main
steamline break with tube ruptures. Okay.
We are talking about leakage rates, you know, a
few GPM may be 100 GPM. That is why these cases look like
repressurizations, okay. In the severe accident analysis, I
keep coming back to saying we will discuss it this
afternoon, we looked at single and multiple tube ruptures,
okay. We have not decided that those should be part of the
design basis. In fact, I think we will never probably
decide those should be part of the design basis, because we
probably don't want those to be likely enough to be
considered part of the design basis. We would like to
preclude tube ruptures, and, certainly, multiple tube
ruptures, given a main steamline break.
The fact that we analyzed them doesn't mean that
we want them in the design basis. You know, we analyze
things beyond the design basis, that is what severe accident
risk analysis is about.
So I think you need to think of this design basis
discussion in the context of relatively small leaks. The
original design basis for a long time was like 1 GPM. Now,
we are talking about 95-05 having cracks open up and, in
fact, probably at very small leakages, but because we can't
really analyze those and assure that the leakages are very
small, you know, we look at them as though they are freespan
cracks and they are not confined and all of that. But these
are still leakages of a few GPM, 10 GPM, 30 GPM, you know,
in some of the more extreme cases, maybe up to 100 GPM, but
none of them looks like a tube rupture.
DR. CATTON: What about tube rupture with a stuck
open relief valve? This is kind of similar, your mild
steamline break where nothing much happens. How different
is it? There you are going to have your 600 GPM and you are
going to have it open to the atmosphere.
MR. HOLAHAN: And, in fact, we analyzed those as
some of the more likely severe accident challenges, but
those are not in the design basis either.
DR. CATTON: You mean the steam generator tube
rupture was an open relief value, was not --
MR. HOLAHAN: It was not in the design basis.
DR. CATTON: But that happened, that has happened.
Didn't it happen at Ginna?
MR. HOLAHAN: No.
SPEAKER: I think they were able to close the
relief value.
MR. HOLAHAN: The main -- the safety valve on the
steam generator leaked for some continued period of time,
but it didn't stick open.
DR. CATTON: Pretty close.
MR. HOLAHAN: I spent three weeks in snowy
Rochester checking out that particular issue in 1982, and
the valve was pretty well seated but leaking.
MR. LONG: A lot of the plants have a requirement
for being able, with a single failure, to prevent overfill
of the steam generator. Now, there is a human error
associated with not succeeding in doing that. So when we
look at the severe accidents, that is included as a
scenario.
It is more a matter of how many tubes do you have
ruptured.
DR. CATTON: No, I understand that. I understand
that. I just thought maybe you were part-way there.
DR. BONACA: Just to complete my thought, however,
since we had it, I agree that there is a design basis issue
and there is a severe accident issue. But I see two
different types of severe accident issues. One is one where
you have a severe accident like a station blackout, and then
you are questioning whether or not the surge line or the
tubes will fail first.
Now, there is nothing the operator can do about
that issue at that point.
MR. HOLAHAN: Well, in fact, --
DR. BONACA: Let me just finish.
MR. HOLAHAN: Go ahead.
DR. BONACA: The other is the scenario where I
have a steamline break, which may happen, and I may have
tubes failing that may be beyond the design basis and I
ignore that. And we are training the operators right now to
operate with ERGs with very specific directions, scenarios
where you have steamline break and tube failures, okay.
There is a full range of analysis being performed behind. I
am trying to understand how credible that is, because this
is a more significant issue in my mind.
We have operators who are now in the control room
trusting that the ERGs will lead them some success under
this kind of condition. That is why I am introducing, I
guess, a third kind of scenario in between, is the one where
you have a design basis moving into a severe accident, but
you have a full body of license documents, I don't know how
licensed the ERGs are, but they are certainly used there,
that at least pretend to be able to cope with those
scenarios.
And that is why I am trying to understand, you
know, as part of this presentation today, how these ERGs can
or cannot be successful.
MR. HOLAHAN: And have analyzed both types of
those issues, both -- what we call the high dry sequences,
core damage leading to tube failure, and, also, what would
start out as a traditional design basis event and then
exceeding the design basis conditions and going to core
damage.
In the context of design basis versus severe
accidents, we call both of those examples severe accident
cases, okay, because you won't find either of them in FSAR.
DR. KRESS: Gary, let's pretend that we were back
in the Dark Ages where all we had was design basis and
didn't have risk and severe accidents, except we kind of had
them in the back of our mind. We defined these design bases
as in terms of probably some perceived frequency at which
they might occur.
MR. HOLAHAN: Yes.
DR. KRESS: But looking at the design basis of,
say, a main steamline break, we had specified in that design
basis, that it leak at the tech spec leak rate.
Now, the reason that specification was in there,
though, was because we had another something in the rules
that said you will not exceed -- cracks that are more than
40 percent throughwall you will plug. Now we are talking
about changing that part of the rules, and we don't have
anything about risk and stuff in there, but we change one
part of the rule, it seems to me like we have already
changed the design basis accident. And you may have changed
it to the point where you might have to talk about changing
the leak rate. And if you change it enough, you might have
to talk about an induced steam generator tube rupture.
It seems to me like we already changed the design
basis accident, and the question is, how much are we going
to change it?
MR. HOLAHAN: Well, I agree that if, in fact, we
were to allow leak rates sufficiently large so that the
events don't look like -- it doesn't look like a main
steamline break, it looks like a much more complicated
event, it looks like a steamline break and a tube rupture,
or it looks like a small LOCA, then, in fact, we would be
having a different event.
But we are not talking about allowing such
leakages. And I don't think we want to go there.
DR. KRESS: Okay. But what I thought was, if you
change the rules about how you deal with the steam generator
tubes, it might very well be that you have no control over
what leakage you are allowing.
MR. HOLAHAN: No, no. I think, in effect, what we
have done is very carefully, in 95-05, looked at the
increased leakage implications associated with change.
DR. KRESS: You say there is now another part of
the rule that does give you a reason to specify a leak rate
as part of the design basis.
MR. HOLAHAN: Yeah. And I think that is part of
what you heard for the last day or so, is that the dose
calculations -- and as early as Jack Hayes' calculations
from yesterday, the dose calculations are done with
substantially higher leak rates for a plant that is using a
95-05 process. But we are not allowing those leak rates to
be sufficiently high that, in fact, they were creating
different accidents.
DR. KRESS: Not allowing them under -- at some
frequency.
MR. HOLAHAN: Not allowing them as expected
results.
DR. KRESS: Expected results.
MR. HOLAHAN: As an expected part of the design
basis.
MR. LONG: You put your finger on one point, and
that is that, initially, there was no understanding of a
difference between the normal operational leakage from the
steam generators and the accident leakage. People weren't
thinking cracks that would open. They were thinking tubes
that could only stand about 10,000 psi, and they might have
pinholes or there would be wastage that you checked and you
patched before it got less than 4,000 psi in strength.
And when we divorced the accident leakage from the
operational leakage, the accident leakage doesn't appear in
the tech specs now, it is a value that is put into the
Chapter 15 analysis. So what is happening is people are
lowering what is in the tech specs, which is the iodine
concentration and the coolant and then through the Chapter
15 analysis, they are increasing the leak rate.
There is no real limit on how far that leak rate
can go. You know, if they are operating at 10 to the minus
4th mikes per cc, and the limit, the assumption is 1, and
the Chapter 15 analysis for 1 GPM, and they are still not at
30 rem to the thyroid in the control room, you know, you can
get the leak rate up to 10,000 GPM and still meet Part 100.
So what Gary is saying is, well, when we grant
these things, we are granting them on a case by case review
and we don't intend to grant something with that high a leak
rate. We have gotten up to 132 in Byron 1 at least, I don't
know about Braidwood, for one cycle, or the last part of one
cycle. I was nervous when we got to 132, and I wanted to
ask, what do we think the real leak rate is if these cracks
are, you know, contained in crud-encrusted tube support
plates? Especially if you shake those tube support plates
with a main steamline break.
We know that the French have done some studies.
You asked about the crud. The French have done some studies
where they have harvested tubes with the support plates
intact, drilled a hole through the support plate, the crud
on the tube, and plugged the support plate. So what they
have is an opening into the crud. And they have
demonstrated it is pretty tight until you move the support
plate with respect to the tube some distance, and then
apparently you crack the crud and you do get some flow. It
is still nothing like the leak rate that you would get if
that hole was in the freespan.
And, in addition, it is a hole you drilled. If it
was a crack and it was essentially in a tube that was being
dented, and that is the reason you had the crack, the crack
may not be able to open and create that hole.
So we don't really have a way of calculating the
leakage as long as that crack remains within the tube
support plate. But we are counting on it being lower than
the value we calculate as if it is in the freespan. And
there is not a very strong knowledge base to tell us how far
we can go in this pseudo leak rate in the Chapter 15
analysis.
CHAIRMAN POWERS: Do you have the description of
these?
MR. LONG: If Emmett Murphy was here, I'd be glad
to say yes, but I'm not sure I know of anybody else in the
audience that has them right now.
We'll try to make sure we get them for you.
MR. HIGGINS: Steve, most of the discussions we've
been having relate to the main steam line break and then
what happens with the possibly-induced leakages.
If you use the stuck-open relief valve as another
initiator, rather than the main steam line break, does the
main steam line break bound that, or do you need to
separately look at the stuck-open steam generator relief?
MR. LONG: When you say bounded, in what sense?
MR. HIGGINS: That you don't need to look at that
and analyze that separately.
MR. LONG: Well, when you get into the accident
sequences and event tree, they're different.
MR. HIGGINS: I'm talking design basis.
MR. LONG: Well, this is what I mean by in what
sense? If you're asking, do you get the same kind of
vibration in the tubes when you use blowdown to a stuck-open
safety valve, I don't think you get that.
The repressurization is slower. We've done it a
few times already.
I would expect that to be a more benign problem
from the standpoint of the blowdown effect.
On the ohter hand, it's something where -- I've
taken the graph down now, but it's something where the
operators have had a tendency to repressurize the system and
increase the delta-P.
MR. HOLAHAN: In the context of the question, the
design basis, the question is, would it produce higher doses
in design basis?
MR. LONG: That's the reason I asked in what
context.
You're saying -- if the question is, would the
doses be higher or lower --
MR. HOLAHAN: Than a main steam line break.
MR. LONG: Probably, I think they would calculate
in as the same in a design basis. I think they'd just
assume that the secondary side is open to the environment.
They would assume that the secondary side is
voided, is depressurized, so there's no scrubbing. And I
would assume they'd get the same answer.
I don't think they have gone into the physics in
any greater detail.
DR. HOPENFELD: Can I make one comment? Is that
okay with you, Steve?
MR. LONG: Sure.
DR. HOPENFELD: I'd just like to put it in
context. And what Mr. Holohan said is very true, that 9505
is limited to very small leakages.
In fact, that was really the main reason why I
converted that DPV to a DPO in July of '94, just before that
9505 went on the street.
And if I remember correctly, I had a discussion
there, and I said, well, anything below 100 or 200 gpm is
not of concern to me, because the operator will take care of
that.
The whole issue was, what we're doing is just as
you describe now, but look what happened. We were 94 and we
basically accepted htat idea that we don't have to go beyond
these small leakages.
And so we have the six years of all that time
that, you know, that we sort of accepted it, and we haven't
-- and that's really the main issue here, why -- I think
we're focusing on it, and that's why 9505 is not adequate.
But we accepted it and let it stay there, and then
we say it is adequate and we're ruling out any leakages
beyond one gpm or ten gmp, and I think you focused the
discussion as to where we should be heading with this.
MR. HOLAHAN: Let me comment on that, because I
think what it says is, the staff's intent is consistent with
Dr. Hopefeld's views; that is, that we both want to keep any
leakages, you know, following a steam line break, to be
small values which can be shown to be things that operators
can handle and are within the dose limits.
It seems to me that the disagreement is with
whetehr, in fact, the thigns that staff has done have
accomplished that goal.
MR. LONG: Yes.
MR. HIGGINS: Related to that, and the operator
actions, as part of the GL 95-05 reviews, were there any
reviews done to see if there were -- that the operators
could still handle the differences in the accident scenarios
between the one gpm leak and now, say, a 100 gpm leak after
the main steam line break, and verifying that the procedures
and the training and so forth were needed -- whether they
needed to be changed or not, or whether any other actions
had to be taken at the sites that are now operating under
these new tech specs?
MR. LONG: Okay, I'm not sure if Joe is going to
get into any of this. He's shaking his head, no.
MR. DONOGHUE: This is Joe Donoghue. I'll be
talking a little bit about this, but the short answer, I
think, is, there were no specific anlayses done fro the
licensing actions. We were depending on the 1477 and the
other analyses that I will talk about.
The conclusoins there convinced us that we didn't
need to do more work on a site-specific basis.
MR. LONG: Were you asking site-specific or just
were there studies done?
MR. HIGGINS: No, whether or not you needed to do
anything site-specific for the plants that were getting
these tech spec amendments, in order to ensure that their
procedures and training were capable of handling these
somewhat different design basis accidents.
MR. LONG: I don't believe we did that.
MR. DONOGHUE: I think the answer, again, is that
some of the anlayses that I will talk about were based on at
least one plant, because that's all we analyzed during the
rulemaking.
We used their procedures as the basis for the
actions and the timing.
That was a very brief synopsis of what was done
but our conclusions were that overall the licensee's
approach here was conservative.
They tried to make sure they were calculating what
would happen -- they used the condition that would give them
the highest peak loads across the tube support plates. To
apply those peak loads to all the tube support plates in the
generator when they took the next step to do the deflection
analysis and they applied a safety factor to those loads
when they did that deflection analysis.
From that we documented in the safety evaluation
that we considered what they had done for this license
amendment was reasonable. However, we made clear that this
was not a generically acceptable approach because of the
limitations MB-2 data. We didn't see this as a basis for a
qualification of this method for generic use.
About six months later I think I was one of the
people here again talking about this license amendment and I
think a subcommittee of the ACRS had some comments about it,
had some additional questions that came up on the ability to
model the flows in the generator during the main steam line
break and we have since used those kinds of questions to
supplement instances where we have addressed licensees
approaching us with this kind of request since then.
I can think of a couple of instances where we have
had very detailed discussions with licensees who have tried
to pick up the methodology that was used for Byron and
Braidwood and we have asked additional questions based on
what we got out of this June meeting and other things that
have come up since then and so far I don't know of any other
licensees that have been able to apply this sort of a
process, this modeling and methodology.
MR. CATTON: One of the problems is that it is a
nonequilibrium behavior. If you think about what happens
before any strong flow starts, the pressure drops, then you
convert to steam and you begin to build up the flow, and
this sort of starts from the bottom to the end so you can
wind up choking and unchoking.
This was the same thing that happens when people
considered the internal loads on the reactor following a
break. You get an expansion wave that travels inside. It's
nonequilibrium. What begins to bring it to a stable process
is when the nonequilibrium process is over and you start
just converting pressure into superheat and to steam and it
is steady.
The loads are going to be quite different. I
think it is the choking and unchoking that is going to get
you, and that is a very quick process at the beginning.
Of course it depends on how many of these area
restrictions you have from one end of this device to the
other, and somehow I was a member of the committee in June
and I don't remember the meeting but I guess if there was
criticism of it, it was probably me.
MR. DONOGHUE: I definitely remember your
questions.
[Laughter.]
MR. DONOGHUE: Scars --
MR. CATTON: It is not clear to me that you can
solve it as essentially an equilibrium process, it's not.
It's nonequilibrium and it's the nonequilibrium effects that
are going to lead to the difficulties.
You have to include them if you want to do it
properly and I don't remember the MB test either. I don't
know what the internals of that thing looked like.
MR. HOPENFELD: Can I just make a comment on that?
We had so many subjects the other day, but I did cover that.
The instrumentation was part of it about the peak
pressure, but that wasn't the main thing.
Remember, I showed you that the volume, the vessel
that was surrounding that slide of tubes, it was a factor of
six or seven higher than the volume occupied by the bundle,
so the whole flow phenomena was controlled but something had
to do with the flow in the tubes, and that was my point,
that you couldn't possibly benchmark RELAP against that kind
of data. It wasn't designed for it.
That was the point and I showed you in the
presentation the volume ratio and I think it is in your
handouts.
MR. DONOGHUE: One thing I remember we did say in
the safety evaluation was that it seemed reasonable to us
that there were so many impediments to pressure waves making
it back to the tube support plates because of equipment that
is in the steam generator that compared to the MB-2 setup we
thought, it seemed reasonable to assume that a lot of those
loads were not going to be any bigger or much bigger or some
phraseology like that than the differential pressures that
were trying to be predicted.
MR. HOPENFELD: There is actually no reason to
assume that.
MR. DONOGHUE: Well, that is assumption we made.
I am just stating what we documented.
I agree, you know, the question about the
equilibrium/nonequilibrium choice for use of RELAP was a big
issue and we --
MR. CATTON: Some of those pressure spikes might
be real. They tried a long time ago with Semiscale, one of
the Semiscale these they begin to get these big oscillations
and they tried to use all of the different codes and they
never could reproduce them.
The problem is when the pressure goes up, you are
condensing. When the pressure goes down you are
evaporating. The thing acts like this huge volume so all of
the frequencies are different. Everything changes.
MR. DONOGHUE: Let me step back for a minute to
again this discussion I tried to say was that -- I may be
able to state it more clearly now -- I am not here to try to
say that we have a basis for resolving the new GSI.
I am just here to say that this is some work that
the Staff is aware of that is connected to the issue and
this is as far as we have gone and we stated in the safety
evaluation there were clear limitations to what we were
doing and why we had problems with this when other licensees
have come in and tried to do this.
The technical details here, the ability to model
these things is certainly an issue and that's why I think we
put all those limitations on this when we first asserted
that.
I have no other information to present on this
topic. Refer back to, I think, material you have in your
truck-load of documents you have -- your safety evaluation
references what the licensee did and the safety evaluation
has the discussion about what the Staff did there and the
things I talked about here.
MR. HOLAHAN: I would just like to remind you that
this relates to something discussed yesterday, that the case
that the Staff approved was one in which because of
uncertainties and other issues we required the licensees
effectively to stake the support plate by tube expansions
above and below it so there was an additional basis for
saying the tube sheet wouldn't move, not just the thermal
hydraulic analysis, so you get an idea of the state of our
comfort and knowledge by the fact that we, even though maybe
your best judgment is that you wouldn't have a problem, we
didn't feel that the analysis without additional actions was
appropriate. Thank you.
MR. DONOGHUE: Yes, I tried to allude to
conservatisms and that is another one that I could have
added to the list.
If there are no other points to discuss, I will go
on to the next issue that I was asked to speak to you about,
which we have talked about to some extent already, how much
leakage can we -- do we think is tolerable during a beyond
design basis, even though I say during design basis
accident, we kind of call them beyond design basis events
here.
This is addressed in Issue 2 of the considerations
document. In there we talk about reports that have come up
repeatedly already and I will just summarize the first one,
NUREG-1477.
We have already talked about that so I won't spend
much time, except to say that there were calculations done
over a range of leak rates, primary-secondary leak rates,
and the conclusion there was that the RWST inventory could
be maintained in accordance, if the operators performed in
accordance with the emergency response guidelines.
The next report, and before I go into detail, I
will just try to put some context on this report, in 1993
when the rulemaking activity was begun, that's when the
Staff were brought together and told to charge off in the
direction of rulemaking, we were challenged by Mr. Thadani,
who was at that time the SSA Division Director, to try to
get a handle on where the risk significance lay here.
This was at the advent. We weren't really
risk-informing as much as we are trying to do today or in as
formal a manner as we are today, but he was very concerned
about these kind of events where we are going, pushing the
envelope or going beyond the design basis line and trying to
understand where we have to focus our attention if we were
going to try to put down a rule to address steam generator
problems.
One of the first things we did was design the INEL
that I think Dr. Bonaca has talked about where we tried to
scope where we thought problems may be.
One of the first things we did was analyze main
steam line breaks with different numbers of tube ruptures.
It was based on -- I will talk about that later -- modeling
assumptions, but the approach anyway was just see if there
is a cliff somewhere that was just outside of the design
basis envelope that we needed to really worry about in terms
of risk to the plant, of risk to the public.
In the end this analysis provided support for us
to concentrate on these -- I will call them severe accident
scenarios but the high and dry sort of things which we ended
up spending a lot of time and effort on in conjunction with
research and produced NUREG 1570.
Efforts continue in that area because of the
uncertainties that we were aware of from that 1570 work.
That was not the end of the process. It continues, but for
our purposes here I am just giving you a context for what
the INEL report represents.
It doesn't maybe go as far as these other efforts
that we call severe accidents. As I said, it summarizes the
analyses with multiple tube ruptures and combined main steam
line break events.
It used the RELAP model, RELAP5 model of Surry and
I think Steve mentioned that as part of this process we
found that we had these same questions about what is the
operator able to or not able to do. The licensee for Surry
was kind enough to send us their complete EOP package, which
the contractor was able to reference and use and they
answered questions for us when we got to the point that I
think Steve mentioned, that we were trying to use a
simulator to understand what operators could or couldn't do.
They answered questions about their own procedures.
In a way it is a very detailed look at one plant
and in a way it is unfortunate because we focused so much on
one plant at the exclusion of other designs but in the
course of the rulemaking we had to concentrate our efforts
somehow and that's what we did.
One issue that I was made aware of that I was
going to spend some time on but I might -- I will get your
sense, Dr. Powers, on whether we wish to spend time on this,
is the assumption that ECCS flow in the event was throttled.
It seems like there's other issues here, but with
the timing I might just jump to rather than discussing the
throttling issue.
CHAIRMAN POWERS: Well, it seems to me that the
critical issue is the kind of time that is available to
recognize and respond to the event --
MR. DONOGHUE: Right.
CHAIRMAN POWERS: -- and start throttling soon
enough. I presume that the operator -- I mean it is safe to
presume that the operator once he starts throttling will
throttle appropriately.
MR. DONOGHUE: Well, that's the question. I will
just touch on it very briefly unless there's questions that
come up.
Just to jump to the conclusions of the report, we arrived at
the point where we thought that, given, given the
procedures, that the RWST inventory at Surrey was sufficient
to handle the combined [inaudible] and multiple tube
ruptures, that dividing line at exact number of tubes is a
point of argument. But it seemed like it was, was not one
tube. It was not even maybe a handful of tubes. It was
probably something a little more than that.
Just briefly on the throttling assumptions,
there's different configurations at different plant, but for
Surrey, you can realign your high-pressure injection through
charging lines and have a throttling capability. The
emergency response guidelines with the CROPs allow -- they
have objectives of maintaining RWST inventory in the case if
you have decreasing steam generator pressure during a tube
rupture. And in order to do that, there are guidelines for
the reduction of injection flow.
Um -- I'll jump to the next-to-last bullet on the
slide. Is that -- the wording in this bullet maybe isn't
the best. But from the range of one to fifteen tubes,
different actions become more important. For the larger
breaks, the number larger number of tubes broke or failing,
it's less important that the operator depressurize because
it's happening already. It's happening by itself.
The other actions that are important are,
obviously, when and how to reduce injection flow and then
the big question -- the biggest question, I think -- is how
and when you get onto RHR. That's what's saving you.
There were people that were involved in this
analysis that still had questions when we got to the point
where we made some conclusions about this. However, as you
heard already, we didn't have information that gave us
credible means for getting to these multiple tube ruptures;
they're very high primary, secondary leakages during the
main steamline break. And we took the direction during this
rule-making trying to develop a technical basis for this of
going off in the other direction that I mentioned before,
the NUREG 1570 analysis.
Talking about the timing, I did bring a couple of
plots that came from the work that was done for the INEL
report. It was also -- the INEL work was also in -- the
NUREG number escapes me. It think it's 6365, steam
generator tube failures, I think is the title. Some of this
common, some of the same analyses ended up in both reports.
The more complete set of analyses were in the INEL report.
And it was kind of a, I guess, a scoping study, a draft sort
of document. It didn't make it into the NUREG stage; it was
a contractor report.
I'll just throw up here -- I might be going
backwards, but, all right, let me do this. If I put up the
one tube-rupture case, it's -- let me see the units. Okay.
With one tube rupture, the RWST inventory is somewhere
around that line. And if you extend, if you extend that
injection rate, that cumulative injection flow up to the
inventory, you can see there is several hours -- I think I
wrote down -- there's several hours that the operators have
to respond.
If you throttle the flow, which is what's done at
about, at about this point, and you throttle the flow, you
get a couple more, several more hours. So the one tube case
seems like there's plenty of time for operators to respond.
If I jump to the very limiting fifteen-tube case, you can
see there where flow was throttling, or without throttling
flow, you can see there's only roughly an hour before you're
done with the RWST.
And I'll point out that for Surrey, there's an
ability to cross-connect to the other RWSTs that's not
included in this analysis. This is just the one thing.
You can see when flow is throttled, that roughly
doubles the time that you have. And that is still of
concern. I wouldn't, I wouldn't feel confident saying to
the operators, given a fifteen-tube -- you know,
double-ended guillotine break of fifteen tubes would be able
to handle things, given that short period of time, even if
flow could be throttled.
I think I have -- here we go. I have a ten-tube
case, which is getting closer to that point that one might
think -- is that clear enough? Yeah -- that one might think
you could survive it. Again, just extrapolating these lines
up to about where the RWST flow, RWST inventory would be,
you can see you get quite a, quite a change in the time that
you have, from about -- oops, that's probably wrong, there
we go -- from maybe a couple of hours to five or six hours,
roughly, which highlights the importance of the operator
actions to reduce flow, but made it apparent to us that it
didn't see, even with this ten-tube failure case, that there
was going to be that -- we weren't on a hairy edge. If
there was just a few hours, we'd still be concerned, as I
mentioned on the fifteen-tube case.
When we were doing, when the INEL was doing this
work for us, these human error probability questions came
up. Steve alluded to some of the efforts that were pursued
to address them. I'm not gonna try to address them here.
I'm not even close to an expert; I'm just aware that that
work was done. However, when we got to a point where we
thought we understood it well enough to get some
risk-informed basis for what we needed to do, the work here
was considered sufficient.
DR. KRESS: What happens to the peak [inaudible]
temperature when you throttle it?
MR. DONOGHUE: Well. I think you just keep the
core covered.
DR. KRESS: -- keep the core covered --
MR. DONOGHUE: Yeah, I mean I have one plot here
where this is fifteen tubes and no operator action, no
throttling. You can see that -- where's that fifteen-tube
case with the throttling on it. You can see that the core
becomes uncovered and you start causing damage. But where
-- yeah, it's well after WST is emptied.
MR. WARD: Excuse me. My name is Len Ward.
There's no challenge to core uncovering.
MR. DONOGHUE: Yeah.
MR. WARD: There are two LIPSI pumps operating in
two [inaudible]. There's a tremendous amount of flow there.
Core uncovering is not a concern unless you have no
injection. And if you have no injection, you don't uncover
until seven hours. And that's because you basically have to
boil off all the fluid above the top of the core, from the
steam generator tube sheet all the way down into the vessel.
Roughly seventy percent of the fluid in the system is above
the top of the core. It takes a long time to boil it off.
If it was flowing out critically, if the break was
in the co-leg, it would lose it a lot faster. So the saving
grace, the good thing about these kinds of events are, the
break's very high in the system and you have to boil fluid
off. And that doesn't challenge injection systems like
critical flow does. So it gives you large amounts of time
before you would start to uncover.
MR. DONOGHUE: Thank you, Dr. Ward. I will just
--
MR. AOPEUFELD: One more comment. There's a
German study showing that only ten tubes, and they were
concerned about turning -- they can't throttle it, so you
have to turn pumps on and off. And they weren't designed
for it.
MR. DONOGHUE: Well, as I mentioned for Surrey,
there's an ability to realign the system to use the charging
lines to, which have the ability to throttle the flow.
In these cases, it was probably a simplifying
assumption that the throttling was done once and we stopped.
I'll just point out that for the fifteen-tube case, the RWST
runs out in a little bit more, around two hours, and that
the boiling is going on for another three to four or five
hours, before you end up having a core damage problem.
As far as -- the throttling assumption I think is
gonna be largely a plant -- it's gonna have to be, it's
gonna have to consider plant differences, design
differences. It's clear. Again, this scoping study, you
use one plant design to see if -- which we thought was
relatively representative of a large portion of the PWRs, to
get an idea of if there was a large risk significance to
these kinds of events.
I think this afternoon, if there's other questions
about the human error probability analysis that was done,
that's the appropriate time to talk about it. Are there any
other questions about that work? Yes?
DR. BONACA: The only reason -- okay, first of
all, I thank you for the presentation. That's the
information I wanted to have.
The reason why I asked directly before is that I
did not see it discussed into the DPR consideration in a
specific fashion, and so I was puzzled and I thought that
you would not be presented that information, which I believe
it's important to our judgments that we have to make here.
And again, I was intrigued by the fact that when we look at
the risk analysis, this information wasn't presented at all.
The DPL consideration. It is discussed under the accident
analysis portion, but it's not considered at all into that.
And that was my reason for asking for that.
MR. HOLAHAN: I would just add that a similar set
of analyses were done about a decade earlier, part of
resolving unresolved safety issues 83, 4 and 5, and had a
similar result for the one-, two- and ten-tube ruptures and
came up with similar conclusions as to the amount of time
available and the likelihood that operators could handle
those cases. And just to simplify again, in the context of
Dr. Hopenfeld's concerns, I think what we're both saying is,
is for a fairly small number of tube ruptures, the operators
have time and can probably handle these.
And in fact, I think both the staff and Dr.
Hopenfeld would say there is a point at which the sequences
do in fact go too fast and the situation is too complicated.
And whether that's ten tubes or twenty tubes, there is in
fact some point at which that occurs. So it seems to me
that the main issue is, what's the likelihood of having
multiple tube ruptures given the steamline break. And the
staff's conclusion is that's very unlikely and we'll discuss
it some more this afternoon, but that same view is not
shared by Dr. Hopenfeld.
DR. BALLENGER: But even if you have fifteen tubes
ruptured, what I just heard was that -- so you run out of
RWST water in two hours, and the operator's completely
flustered and can't deal with it. You've got six hours more
before the core is uncovered.
MR. HOLAHAN: No.
DR. BALLENGER: No?
MR. DONOGHUE: That's a total of about six hours.
You have maybe three or four.
DR. BALLENGER: So you've got four hours.
@@ DR. BALLENGER: Three or four hours.
@@ DR. BALLENGER: You've got three or four hours
more grace period, if you will.
MR. DONOGHUE: Yes.
CHAIRMAN POWERS: The problem is, once you concede
fifteen, you've got to concede twenty. Once you concede
twenty, you've got to concede twenty-five. I mean, one or
two is different from fifteen.
MR. HOLAHAN: And I think that as the number of
tubes would increase, in fact that amount of time available
would decrease, because it wouldn't just be boil-off. You
could actually have a system blowdown if the number of
failed tubes was large.
MR. DONOGHUE: If there's no further remarks about
that, I'll go to my last topic, which I won't even try to
say is going to be brief, even though I only have a couple
of slides.
We've touched on this I think earlier, the leakage
that could develop during a depressurization event. And
just point out that I have one other page that I want to
make sure you have. It's a list of events I'll get to in a
second.
When we talked about this in the DPOP
considerations document -- this I think is issue 2. And let
me see, break leakage. I think we've mentioned that there
have been depressurization events, we have not seen primary,
secondary leakage associated with those events. Those kind
of events are usually association with stuck-open relief
valves, loss of feedwater, or some combination of those,
those kind of failures. And when we look at the reports for
those, some of those events -- which I'll show you the list
in a second that I'm talking about -- we don't see a
discussion about primary and secondary leakage.
If there was primary and secondary leakage, there
are steps in the procedures that the operators use that take
that into account. If there's contamination going to the
secondary side, there's certain things they need to do.
They need to monitor for it, but there's also steps to take
to, to try to limit the contamination. But what's
important, I think, is that for the events that we're aware
of, that when plants returned to power, there was not tube
leakage that was reported to the NRC. We didn't see tube
failures manifested in leakage from these type of events.
DR. SIEBER: Could I ask a question.
MR. DONOGHUE: Yes.
DR. SIEBER: [inaudible] had a blowdown during a
[inaudible]. Was there an inspection or do you have any
information related to the condition of that steam generator
prior to its being put in service?
MR. STROSNIDER: Yeah, this is Jack Strosnider of
the Staff. You're referring to an event that we heard about
the day before yesterday, I guess. We've asked the staff to
go look at the docket and see if we have anything reported
to NRC. I can't tell you the answer at this point, but we
are pursuing that question.
MR. DONOGHUE: I would just add that I think in
the documents that you have, there's accounts of those
events, of that and I think and event at Robinson. And just
looking at those accounts, I didn't see any discussion about
the -- you know, going back and looking at the steam
generator. I did look at some other information I think
EPRI had on repair histories for tube. And I'm not sure if
Jack's staff has looked into that. I'm sure they are.
But, you know, it didn't seem like there were
tubes repaired at -- that's just speculation on my part.
That's just basically absent information; I think Jack's
staff will be able to answer that better. But for these
events, these are just examples of the type events that we
mentioned in the DPO considerations document, where in some
cases and in one case here, both steam generators lost
inventory.
The primary pressure changed, but the primary did
not depressurize during these blowdowns and there was
significant differential pressure across the tubes. I
wouldn't call these type of events are gonna produce any
kind of dynamic events that would be something, you know,
that could help address the new GSI. These are just purely
instances where you have a high differential pressure across
the tubes. But look at the LERS across these events, or in
the case of [inaudible], there's a detailed NUREG on that --
MR. HOLAHAN: My recollection is sitting up all
night in the operations center watching the Davis-Bessy
cooldown. And I think we asked some questions about whether
there was any leakage. And I think they did not have a
problem.
MR. DONOGHUE: Thank you. So there's another
piece of information that's helping. We didn't see tube
leakage after such events. And after we -- we don't know of
this being a problem for depressurization events. Which
leads to some assurance that if there's a high differential
pressure across the tubes, even after some period of
operation, that we're not going to see leakage develop.
HOPENFELD: [OFF MIKE]
MR. DONOGHUE: Say again?
HOPENFELD: [OFF MIKE]
MR. DONOGHUE: I'm not sure --
MR. STROSNIDER: That's correct. None of those,
none of those units particularly, you know, had the generic
letter number 505 alternate repair criteria in place.
HOPENFELD: [OFF MIKE]
MR. DONOGHUE: No, but the point is that the
plants operated -- these plants did have some tube repair,
although they weren't ultimate repair criteria. And it's
just showing that we don't have information to show us that
the tubes are going to leak excessively after a high, high
differential pressure is applied to the tubes.
CHAIRMAN POWERS: I feel, I feel absolutely
obligated to point out that after 24 launches of the
shuttle, we didn't have any evidence that we'd have a
Challenger.
MR. DONOGHUE: I understand --
CHAIRMAN POWERS: Small databases are useful for
contradicting hypotheses, not confirming them.
MR. DONOGHUE: I see. Well, we just wanted to
present the operational information that we knew about when
we wrote the DPO considerations document.
CHAIRMAN POWERS: And I think that's, that's what
the Committee asked for and it's useful.
MR. DONOGHUE: Unless there's other questions or
remarks, that's all I have for today.
CHAIRMAN POWERS: Are there other questions?
Speaker? Seeing none, I'd like to pose a question to Dr.
Shack. We let him get away way too easy on this, on his
presentation. So we'll take just a minute or two. Dr.
Shack, if I ask you a question, may I suggest you just sit
here in the designated Federal officials' seat.
[LAUGHTER]
CHAIRMAN POWERS: When you spoke, you spoke of
both circumferential and axial cracks and presented a
mind-numbing amount of data and analyses that suggest that
we really understand circumferential and axial cracks in a
fair amount of detail. But I'm reminded of the cracks that
people show us that show that they are not completely
circumferential or axial in all cases. And I'm wondering
how -- do we have guidelines to tell us how to apply all
that knowledge to more realistic cracks that have some
convolution of shape that might suggest that they have some
circumferential characteristics and some axial
characteristics.
DR. SHACK: One moment. I suspect this is more a
question of what the regulator is done when he's faced with
those questions. I think most of the time, unless one has
better information, one makes a rather conservative bounding
projection of the cracks, so that you're, you're almost
collapsing cracks that are separated by ligaments onto a
plane and using that kind of bounding analysis. There are
rules in the code, you know, if you could clearly
demonstrate the separation of these segments, but in many
cases, the resolution of the NDE isn't good enough, so you
would end up collapsing them.
We're looking in a research basis at what you do
when you have a combination of circumferential and axial
cracks together. We -- I think that one would again take
rather conservative estimates of how that would work, by
projecting everything onto a single plane.
SPEAKER: Cosine of the angle?
DR. CATTON: You don't get an oblique crack?
SHA: There's thousands of cracks out there, you
know. Never's a long time.
[LAUGHTER]
SHA: I would say that most of the time, you get
circumferential and axial cracks. I sort of explained it to
Dana, that one of the things you seldom see -- stress
corrosion cracks don't grow under pure sheer, typically.
We've tried to grow them under torsion, which is one way to
get a pure sheer state. You seem to need normal stresses,
and so they line up along principal stress axes, which in a
tube happen to be -- it's this way.
Now again, at a roll transition, the stress state
is always a little more complicated and things might not be
so simple. But the stress patterns there are so
complicated, my guess is you end up assuming that they're
projected into some 360-degree plane and you'll, you'll find
that you don't want to live with the results.
Again, if it was a small crack at an angle, as my
results show, you know, the 300 degrees, you're not gonna
quibble over one short, small crack. But if you have an
extensive amount of cracking -- but again, it's the
regulators who actually handle that problem.
MR. STROSNIDER: Yeah. This is Jack Strosnider.
And just to follow up on what Dr. Shack indicated,
typically, typically because of principal stresses in the
pressure, the large hoop stress and that, you're gonna see
axial cracks or circumferential cracks. That's -- in order
to get something that's, that's at some sort of angle, you
need something like the ODSCC under the tube support plates,
which is -- and even there, it, it lends itself to the
creation of something closer to intergranular attack, as we
said earlier. So you've got a network of cracks.
But as that crack develops -- and this is what's
applicable under generic letter 9505 -- is you leave, if
it's left in there long enough and as it develops, it will
develop a principally, an axially oriented, dominant crack.
You know, the one area where -- and unfortunately
I don't have any of the staff here who can go into a lot of
detail on this -- but my recollection is that down in the
crevice of the tube sheet, you know, some of these plants,
the tube are not full-depth expanded into that [inaudible]
tube sheet. It might be expanded three inches, and then
you've got this 20 or 21 inch crevice.
Down in that crevice region, where we have some
alternate repair criteria, all right, which are based on
depth into the, down into the tubesheets and the inability
to pull it out, friction loads and that sort of thing, my
recollection is that we have placed on some of those repair
criteria, some requirements that when they do the
inspection, they verify that the cracks do not have a
significant circumferential portion, because they do
sometimes grow. In that sort of environment, they may not
grow at perfectly along the length of the tube. So we did
establish some criteria there. That's my recollection. If
you want more specifics on that, I'll have to get it for
you. But it's something like that --
CHAIRMAN POWERS: Yes, I'd like to --
MR. STROSNIDER: -- it's that sort of unique sort
of situation down in that crevice where you might tend to
see something, you know, like you're --
CHAIRMAN POWERS: Yes, I'd like to know why you,
you asked for something particular about the circumferential
character to the cracks, because I got the distinct
impression from the explicit words that you could tolerate
circumferential cracks a lot better than you could axial
cracks.
MR. STROSNIDER: Well, I'm, I'm not sure if I
follow everything. Your question exactly is -- it was
discussed this morning, circumferential cracks are more
tolerable in the sense that you have, they have lower
stresses on them trying to pull them apart. In reality,
when you look at circumferential cracks, at least at the top
of the tubesheet, they tend to have a lot of ligaments in
them. That may not necessarily be true of some of the
primary water cracks that, that show up. But it's largely
because of the lower stresses that are on 'em.
With regard to these criteria down in the
tubesheet, as I said, we have to get some more specifics for
you. But sort of a general concept is that the ideas, you
want to make sure the thing doesn't pull out, all right?
And if you had the potential to grow a circumferential crack
completely around the tube, then, you know, that, that, you
know, too close to the top of the tubesheet, then you might
have some concern about whether you've got enough tube down
into the tubesheet to keep it in place.
All right. The other aspect of it is that the
axial cracks that are in there, you know, you have to look
at them from some sort of leakage point of view.
So I don't know if -- does that address your
question?
CHAIRMAN POWERS: Well, I mean, the question is
enormously naÃve. You asked for some special NDA activities
in that region for circumferential cracks. I just wanted to
know why.
MR. STROSNIDER: Well, and I in general the
concern is that, as I indicated, you know, there's criteria
with how high are -- I don't know which way to describe it.
It's a better way to picture, you know. You don't want
degradation too close to the top of the tube sheet because
you want enough of the tube down into the tubesheet to
provide the restraint. And also, you don't want
circumferential cracking because, because it could impact,
you lose some of the frictional load and stuff.
CHAIRMAN POWERS: I guess I could understand the
frictional load. But my, what I understood you to say was
that you asked for more NDE in this region because the tubes
weren't full expanded and you had a long crevice region,
which isn't holding the tube in except for whatever friction
there is.
MR. STROSNIDER: And it's probably best if I get
one of the staff to come provide you some detail. But my
recollection is that what happens is that you can allow some
axial cracks, you know, getting closer to the top of the
tube sheet because they're not gonna impact at the pull-out.
All right, but if those axial cracks start showing some
circumferential orientation, all right, then you want to
limit that.
But let me make a note to get some more detail on
that for you.
CHAIRMAN POWERS: Yeah, I guess I'd like to
understand a little better because I came away distinctly
with the impression that circumferential cracks were rare,
even though you might have a limited capability to detect
them, that they were just much more tolerable.
DR. SHACK: No. I didn't say that, I don't think.
They're not rare.
CHAIRMAN POWERS: You didn't say that. I got that
from another speaker.
MR. STROSNIDER: I would point out --
CHAIRMAN POWERS: I could be equally wrong.
[LAUGHTER]
MR. STROSNIDER: I would point out one other
thing. And we didn't get into a large discussion about
leak-before-break in steam generator tubes, all right. And
in general, we do not credit leak-before-break in steam
generator tubes because, obviously, we've had failures where
leakage either wasn't there or wasn't adequate for the
operators to head off the failure. So we, in general, we
don't credit it. However, if you look at all those leakage
events that have occurred, in a large majority of the cases,
it in fact does come into play.
For circumferential cracks, particularly at the
top of the tube sheet, as we indicated this morning, the
failure loads or stresses required on those are 7,000 to
8,000 psi. I mean, they're, they're not a like a brand-new
tube, but they're still pretty high. However, they, they
may develop a leak, and that's an area where, you know, you
could argue that leak-before-break is the most likely
failure mode for those. All right, so -- they are somewhat
more tolerant from that regard.
MR. DONOGHUE: Nothing else for me.
CHAIRMAN POWERS: Okay. Thank you. I think we're
in a position now that we can take a recess for lunch until
1 o'clock.
[LUNCH RECESS 12:10 P.M. UNTIL 1:00 P.M.]
. AFTERNOON SESSION
[1:00 p.m.]
CHAIRMAN POWERS: So, I'll note for this portion
of the meeting Mr. Dudley is acting as our designated
federal official, guiding me with an iron hand, right?
I think at this point we are scheduled to turn to
the simple and easily tractable issue of severe accidents.
And Mr. Long has drawn the short straw here.
MR. LONG: That was a much shorter introduction
than I expected.
CHAIRMAN POWERS: Well, it didn't look like any of
your compatriots are here to help you either.
MR. LONG: Okay. We have a few of us on here.
I've got the first few on severe accidents. And we went
over a little bit of this earlier. Severe accidents are
pretty much the things that we were talking about earlier
that are starting from design basis accidents, but becoming
more complicated and perhaps going towards core damage.
Plus things that we've always analyzed as if they
were going toward core damage. So I've tried to put a list
of them up here.
CHAIRMAN POWERS: Let me ask you this question.
Would I be completely adrift if I argued that I can tell
operationally whether an accident is severe or an accident
is design basis by looking where the operators are working
from the emergency procedure guidelines or working from the
severe management action plan?
MR. LONG: Well, certainly they're going to start
with EOPs before they get to the severe accident management
plan, anyway. So it's more a matter of how far does the
thing progress. Things like steam generator tube rupture
that are supposed to be design basis accidents still have
the possibility of becoming complicated or adding errors
committed that will take them all the way to core damage.
So there's--
CHAIRMAN POWERS: But I can tell the difference
between a design basis steam generator tube rupture accident
and a severe accident involving a steam generator tube
rupture. If I wait long enough by seeing if the guy stays
and his EPGs or goes to Sam?
MR. HOLAHAN: That the procedures are written in
such a way that the event, you know, drives you through the
process.
CHAIRMAN POWERS: Yes, I understand that.
MR. LONG: I guess what I'd say to try to answer
your question is the design basis accident is one that
pretty much doesn't go beyond what chapter 15 analyzed.
MR. KRESS: I think that's a best way to do it.
MR. LONG: And a severe accident is one that has
gone somewhat past what chapter 15 analyzed that you now
have core damage that's significant enough to make
substantial radiological release from the core. And there's
probably a substantial gap in between those two, where the
accident is probably beyond design basis, but not yet
severe.
At any rate, we try to break them down into
sequences that start along those paths, and just to sort of
get the group of things on the table that we're going to
talk about. There's the spontaneous tube rupture that start
as a design basis accident at least, and perhaps gets as far
as core damage.
There are sequences initiated by things that are
within the design basis, like secondary depressurizations
that increase tube pressure differentials and may lead to
things beyond design basis, like rupture of tubes or leakage
beyond the design basis values, that also have a potential
for getting to core damage.
There are sequences like ATWES that are not really
design basis, also increase tube differential pressure by
increasing primary system pressure rather than decreasing
secondary side and we do have in PRAs going to core damage.
And then there are the things that don't really
involved tube rupture to get you to core damage, but -- such
as station blackout or other things that usually loss of
second cooling, loss of primary inventory, that may, in the
process of getting to core damage, also affect the steam
generator tubes and perhaps convert some of these things
from accidents where the core damage is contained within the
containment structure to accidents where the radioactive
materials bypass the containment structure.
So all of these would potentially increase risk to
the public in terms of radiation exposure and the health
consequences that are created by that.
So that's the--
CHAIRMAN POWERS: You said increase, do you really
mean contribute to?
MR. LONG: They increase the probability of is the
best way of saying it.
MR. KRESS: They contribute to, Dana, but I think
the increase would be comparing to what you would have if
you didn't have the alternate repair criteria.
MR. LONG: I'm not sure what his baseline was. In
other words, there's the--
CHAIRMAN POWERS: I wasn't either.
MR. LONG: There's the baseline LERF, which is
just taken as what people are calculating in their PRA, and
that's not necessarily a complete representation of LERF.
We think there's some pieces missing.
CHAIRMAN POWERS: Gee, I can't imagine what they'd
be.
MR. LONG: Well, we'll talk about them. And then
there's the question of if you have a licensing amendment
request, would that change you from whatever the baseline
was. Maybe that baseline had to be augmented to begin with
to something that's substantially higher. Then we'll get to
the last topic of the day, which is the risk informed
decision process.
Anyway, to just launch into the different
sequences. The spontaneous tube rupture sequence is one
that's been in PRA's basically since the -- I guess it was
Point Beach Plant, pointed out that they needed to be in
PRAs. I don't believe it was in WASH-1400. And I wanted to
start with this one because it's the one that's been
analyzed most so far, and we can learn some things from it.
It's been treated in all the IPEs. Most of them
have a number that's very close to one times ten to the
minus two as the initiating event, frequency per year.
However, they have a wide range of results in the core
damage frequency in a per year basis that results from that.
We don't fully understand the reasons for the wide variety
in results. But we look into it -- we see that for the
results that come up on the high side, they seem to be
dominated by human error probabilities. And for the results
that come up on the low side, there seem to be more hardware
failures and less human failure represented in the dominant
cut sets.
MR. CATTON: Are the ones that are on the high
side from plants where they've had an event?
MR. LONG: Not necessarily.
MR. CATTON: Not necessarily?
MR. LONG: So it looks like the human error
probability modeling process is really what creates a lot of
the difference in the results we see in the IPEs. And it's
not just a matter of what number they put on the human error
probability that appears in the model. It's also where they
put the human error probabilities in the model--which ones
are represented, which ones may not be represented.
At any rate, as modeled now, especially if it's
the batch that have the higher values for a steam generator
tube rupture core damage contribution, that's usually one of
the dominant contributors to the total results in public
health consequences, not the core damage, but to things like
LERF-50 go to LERF, but more to population dose, cancer
effects, other effects from that population dose.
So that makes it pretty important to understand
how that would be affected by licensing actions.
Moving on to the next one.
MR. HIGGINS: Phil, so does the licensing action
for 9505 change the -- either accident sequence frequency
core damage frequency or LERF related to this type of severe
accident?
MR. LONG: We don't expect anything we're doing to
increase the probability of spontaneous tube rupture --
anything we're approving. We're trying to keep things that
we approve to where they would still meet the three delta p
criterion, and you know, the intent is to not increase this
probability.
Sort of going in the order that the questions were
asked, but not quite. One of the questions was about
station blackout accidents, and really what we're talking
about is a core damage frequency, the component of core
damage frequency that has high RCS pressure and a dry
secondary side. In other words, the high dry frequency as
we call it. And when we say high, we don't mean necessarily
sitting on the safety valve set point, but down to where you
still haven't really injected your accumulators.
There are a lot of ways of getting there. It's
usually mostly station blackout, but some plants are
actually dominated by loss of DC bus or buses. Small LOCAs
with loss of secondary cooling. Pretty much anything that
allows the core to become uncovered and damaged and has the
secondary side dry has the potential for producing a
transport of heat to the tubes without the secondary fluid
to cool the tubes.
The concerns then for steam generator tube rupture
affecting the progress of this event are the -- if a loss of
secondary integrity occurs to the point that the secondary
depressurizes as well as dries out, you have a delta p just
like you would for the main steam line break, you can -- if
you can rupture or cause gross leakage in tubes for the main
steam line break event, you could also for some of these
sequences.
The other aspect is if the tubes are strong enough
to withstand the delta p at normal operating temperatures,
if they become hotter, they may still fail as the, you know,
material weakens at the higher temperatures.
So these perturbations by tube degradation are
usually not included in the IPEs. The -- some of the IPEs
have picked up the one point four percent, I think it is,
that was in the NUREG 1150, 4550 plant risk models as an
expert elicitation process for what fraction of the time
they thought there would probably be a pre-existing tube
flaw that would be sufficient to cause the tubes to fail
first under these conditions. But most plants haven't
picked up a -- anything beyond just that one number that
came out of expert elicitation and looked at the sensitivity
of that number to tube integrity measures that are plant
specific.
When we did NUREG 1150, we tried to go into a
couple of PRAs, primarily the Surrey NUREG 1150 PRA, and
search for high dry sequences and try to get some estimate
of the timing to see if the RCS would be pressurized before
the steam generator or the other way around, and asked
ourselves if we'd have a -- just a pressure-induced failure.
I mean, a lot of work was done to see if the failures could
be thermally induced.
The factors that we had to consider for RCS
pressure involved the reactor coolant pump seals and the
burn off rate they would have. At least early on, there
were a lot of large seal leak scenarios that were
considered. Also, if the tubes are leaking substantially,
that's another issue with -- that I'll get into a little
deeper later, but it has to do with RCS pressure at least.
Pressurizer valves may also stick partially open. Avery did
some studies to determine if you continued to pass either
water or hot -- they didn't look at very hot steam -- but
repetitive openings of valves has a tendency to cause the
valves to not fully close, and we've run some cases where we
stuck pressurizer valves partially open.
These all seem to have effects that are reasonably
important to consider, but they're complicated. So we'll
get into a little more of that later.
Other things to consider are what's happening on
the secondary side, the main steam line safety valves may
stick. That's been in a lot of the PRAs for a long time.
There are other valves, like the turbine driven aux feed
water supply -- steam supply line. If you run out of
batteries, you may need to -- and can't control the
governor, you may need to think about closing that line.
The MSIVs may leak. We've had events where plants have been
able to, you know, lose a lot of fluid through an MSIV, and
not really notice it during their normal operation for start
up. The time they seem to discover this is when they have
to do a secondary site hydro, and they realize that they
can't pressurize the secondary site for the hydro, and then
they go find the leak.
The thermodynamics of the reactor coolant system
heat up control how the heat can be transported from the
oxidizing core, really, is the point at which this is
important, after the core has been uncovered and heated up
on decay heat, and it starts to get a chemical addition to
the heat up rate due to the oxidation of the clad is about
when these things really start to become important for over
temperature in the tubes.
And the thermal hydraulics of the process can be
important here if the -- if there's full loop convective
circulation through the tubes, especially if the tubes are
depressurized on the secondary side, it looks like even new
tubes, pristine tubes, will not be able to withstand the
heat up without being the first component to fail. They're
thin, and the heat up process is fairly rapid.
However, there are places where this full loop
circulation can be blocked. There can be water in the
suction to the reactor coolant pump, and they call it the
loop seal. There can also be water left below the core
that's blocking the down comer such that you don't really
have a path below the down comer and up through the core.
If you get into scenarios where you are
depressurizing -- excuse me, drying out the core without
depressurizing, and then you put a fairly small leak in the
system, you have a potential for boiling away loop seals,
getting flows re-establishing loop seals. And this has
turned out to be quite complicated. So we've had -- you
asked some questions about stylized sequences, and in NUREG
1150 time period, we were looking at either the reactor
coolant system stays at the safety valve or pressurizer for
set points, or there were large leaks in the RCS, in the
reactor coolant pump seals, and the whole thing
depressurizes; the accumulators eject.
We've more recently started to look at situations
where leakage is in one or another part of the RCS -- take
the pressure down low enough to stop cycling the pores, but
not necessarily to dump the accumulators or at least not low
enough to really remove the pressure from the tubes
completely. And this prolongs the accident. It gets into
much more complicated phenomena.
There may be different delta p's among the
generators, and within the generators the temperature may
vary. And I have a slide I just want to throw up here real
quick. But this is from the 1/7th scale test. I don't know
if this is a slide you'll be able to read. At any rate, the
-- there's 216 tubes, and the sort have been partitioned
along this dotted line to be the tubes that were thought to
carry flow upward and -- well, out of the inlet plenum or
the outlet plenum, and then the rest of the tubes -- the
outlet plenum being over here -- and then the rest of these
tubes were carrying flow from here back around into this
side. And the temperature distribution on here, although
the peak is right around in here, and you can it's, in this
case, 178 roughly degrees. Over here, it's maybe 143
degrees. Over here, it's a 145 degrees. So there's quite a
variation in this batch of tubes that's modeled as being the
hot bundle.
And when we get into trying to act -- ask
ourselves how big does the crack have to be to cause failure
if experiencing temperatures in the hot bundle. Right now,
we don't have a good way of describing the distribution of
temperature within that hot bundle. RELAP gives us one
number for the entire hot bundle. So that's been a problem
for us as well.
So all of this has gone into our thought process
about how to deal with the high dry frequency. At this
point, we're going to have discussions later on how to model
that. I don't want to go into it much more deeply yet.
We'll get back to how we used it later. Are there any
questions on the high dry sequences as to what they cover or
what we're trying to include?
MR. HIGGINS: Are you going to get to results in
terms of numbers, increase in delta LERF, and that sort of
thing for 9505?
MR. LONG: 95-05 we wouldn't expect to see any
delta LERF. That was part of the premise, that we weren't
going to be increasing core damage frequency or LERF.
MR. HIGGINS: But wouldn't the times at which you
get two failures change with the 95-05 criteria, so why
wouldn't you see a difference?
MR. LONG: Okay. In the -- 95-05 allows
degradation to occur where it's confined by tube support
lights. And the thinking in the time was the blow down that
you would get from the secondary side, from the things I was
discussing -- are stuck valves, other things that we've seen
blow downs before. We don't expect that to really move the
tube support plates off of those damaged portions of the
tubes. There is one issue that I don't think anybody has
really--
MR. HIGGINS: But I didn't think you were taking
credit for the tube support plate restricting and preventing
leakage.
MR. LONG: Those are steam line break cases. In
the design basis analysis for steam line breaks, they are
not taking credit for it. In a risk assessment, we would be
taking credit for it, unless we had a reason to believe it
wouldn't be there.
MR. CATTON: There's another factor, too. You
know, even if -- although we may disagree on the mixing and
so forth, the cracks that are going to be in the vicinity of
the support plate, you got a huge heat sink. So that's
really not where you're going to heat up the tubes.
MR. LONG: On the support plate, I'm not sure how
big the heat sink turns out to be. But--
MR. CATTON: Well, but it is a heat sink.
MR. LONG: To some degree, yes.
MR. CATTON: So if you're going to heat anything
up, you're going to heat the freestanding parts of the tubes
more than you're going to heat the -- where there's a big
solid chunk of metal.
MR. LONG: You're talking about the support plate,
and not the sheet. The sheet clearly is a big solid chunk
of metal.
MR. CATTON: About two feet thick. Even when you
have three-quarter inch thick, that's steel, and it's a heat
sink.
MR. LONG: If it's there, it will prevent.--
MR. CATTON: It will prevent it from heating up as
fast as it goes somewhere else.
MR. LONG: The one thing to think about, though,
is if you have a bundle of tubes that are hot, and these are
hotter than those, what does that do in differential
expansion? We think it will probably kind of bowed the
tubes, but we haven't really looked at if that effect on the
tube support plates.
MR. CATTON: Are those the thermal couples that
were in the inlet of the tubes, or are they just below the--
MR. LONG: I believe, if you look up here, these
are -- the dots are one inch from the tube sheet bottom.
The closed triangles are three inches from the tube sheet
bottom. I believe that's in the tubes. And then the open
pointed down triangles are point seven five inch below the
tube sheet. So there's a variety of them in here.
MR. CATTON: Surely, you'll explain all this
later.
MR. HIGGINS: So, Steve, in this last group you've
included both ones that would induce tube rupture by both
thermal high temperatures on the primary side, due to the
core damage and ones that would be induced due to the high
delta p?
MR. LONG: Right.
MR. HIGGINS: And you're saying neither of those
would result in increased numbers for the 95-05?
MR. LONG: When we did 1570, we weren't really
thinking about the 95-05 degradation. We were thinking
about free spanning cracks that were in the sludge pile or
other areas that were not confined. And the -- at the time
that we did NUREG 1570, the industry had asked for
essentially a five percent conditional failure probability
in the free span for main steam line break, because they had
looked at NUREG 0844, and NUREG 0844 had concluded that
basically we wouldn't back fit them if we had five percent
conditional failure probability of tubes, given steam line
break. So the industry was sort of turning this -- well,
it's not bad enough to back fit them into a performance
criterion if they could. We were trying, in NUREG 1570, to
add to our knowledge base what would happen if we had that
level of degradation. So the numbers in 1570 are -- we
tried to peg to something that would give a five percent
conditional failure probability for steam line break, and
then ask ourselves for that, what do we expect in severe
accident conditions.
So we weren't trying to develop something we would
accept. We were trying to explore what would be the case if
this occurred.
MR. HIGGINS: Yeah, I guess what I'm trying to
explore and see -- and I thought maybe we would get to it as
we went through this was that whether or not for the various
different types of core damage sequences that are important
to steam generator tube rupture in 95-05, what do the
results look like in terms of increases in core damage
frequency or increase in LERF, and is it reasonable in
severe accident space? Is what you've done reasonable in
severe accident space? But it sounds like what you're
saying is you don't have all those numbers?
MR. LONG: What I'm saying is when we did 95-05,
actually when we did the interim plugging criteria, which
became 95-05, the intent was not to increase core damage
frequency or LERF at all. And the basis for that was the
belief that we had confinement of the damaged area, the
degraded area, by the tube support plates. There was a
concern for what was considered to be a very hypothetical
kind of main stream line break that you might move the
support plate relative to that degradation. We didn't know
how to calculate the -- actually the clamping effect of the
tubes on the support plates, given that the -- you know,
that the -- or I should say it the other way around. But
the support plates are denning the tubes. That's why we
have the degradation, and there is quite a force per tube,
which they have to overcome to pull them.
So there was a feeling that realistically the
support plates were pretty well held in place by the tubes.
From the design basis, they were having trouble
quantitatively crediting it, so they did not. And they went
to the -- what we were considering to be a hypothetical leak
rate and a hypothetical conditional burst probability. Now
there is some stuff in 1477--
MR. HIGGINS: Are you going to represent anything
to -- or do we have anything already that provides some
justification for that?
MR. LONG: Provides justification for the tube
support plate not moving?
MR. HIGGINS: For clamping the leakage, any
leakage that might come from failures or to prevent failures
of those defects?
MR. LONG: Where's Joe?
MR. DONOGHUE: I'm sorry. Which one?
MR. LONG: The question is, are we going to
present anything about reason to believe that the tube
support plates really are held in place as far as doing a
risk assessment is concerned?
MR. DONOGHUE: I don't have any material on that
one.
MR. LONG: Yeah, I don't -- I guess that's
something that we could take as an action item to try to put
together.
MR. HIGGINS: But you're saying that's the basis
for your concluding that none of these core damage sequences
show any increased in LERF?
MR. LONG: That was the basis for granting the --
you know, the 95-05 plugging criteria. Without having done
a detailed study of the leak rate in a realistic framework.
In other words, when it came to what I would put in a risk
assessment, the risk assessment was not done until after
those were evaluated. We had not done 1570 when the interim
plugging criteria was out. We had discussed it. There was
a qualitative feeling that for a realistic blow down, the
tube support plates would be in place, and that's really
what we were basing it on. The risk -- that -- the --
that's what we were basing the lack of a formal risk
calculation on at that time.
MR. BONACA: Let me just ask a question. I know I
understand what happened at the time before 1570 and 1477,
but my main concern here is -- the thrust of my question was
because the DPO, the DPO claims that a certain scenario
which you define severe, it's possible. It is likely, and
they assign a high risk frequency to it. And that's the
DPO. When I read the DPO consideration, I found that the
very scenario that they are discussing there is not being
quantified or addressed in the DPO consideration. The DPO
consideration only addresses the possibility of leakage
considering 1477 up to about a thousand GPM, with some
assignment of risk to that -- ten to the minus six. And
then addresses the consequential failures of tubes resulting
from station blackout, and then it says any more, you know,
tube failures from main steam line break is not considered
likely or possible. Therefore, there is not quantification
or that. So it's very hard to evaluate the DPO
consideration because there is a lack of information
regarding how -- because it's the only denial that the event
can happen. And just the point I want to make is that --
that's why, by the way, to explain it, I jumped to the INEL
report, because the INEL report does also some human
reliability analysis. Now, the reason for digging into it
for me was to understand how reliable the reliability
analysis was, and I'm beginning to get a sense of it now.
MR. LONG: Okay, let me try to tease apart two
things.
MR. BONACA: Yes.
MR. LONG: Jim has been asking about 95-05, and
you've been asking about the DPO. And they're not
identical. What we were trying to do were the NUREG 1570
work was think about things that we thought might be able to
fail in the free span. In other words, another part of the
DPO was that there are so many cracks out there we can't
detect -- that might be in the free span, not just things
that are detected but left in service under the support
plates -- that, in fact, you might get a huge leak rate, if
not actually ruptures, in the free span. So when I was
answering Jim Higgins' question about 95-05, I wasn't trying
to say we didn't consider other reasons that there might be
failures of tubes. The -- but -- now to go to the other
reasons.
I mentioned earlier that we received from Dominion
Engineering some estimates of populations of flaws in steam
generators, in various types of degradations. And the ones
we concentrated on for NUREG 1570 were the ones that were in
the free span, not the ones that were under the support
plates. And we tried to pick a distribution of those which
turned out to be either average distribution that looked
like it would give a conditional failure probability under
main steam line break of five percent. And that was
basically one tube out of five percent, one or more. But
the way it works out is essentially one.
With 1570, there was some other work done in
parallel with that to ask, and I'll try to get to that at
least the beginning of that in a minute here, to ask what
are the thermohydraulics for a larger number of tubes. But
when it came to the risk calculation, I need the initiating
event frequency, which would be something like the
non-benign -- non -- what was the word you were using
earlier this morning? It was the gentle main steam line
break or something?
CHAIRMAN POWERS: No. It was Ivan's mild steam
line break.
MR. LONG: Mild main stream line break. So the
wild and wooly main steam line break frequency times some
conditional probability given that wild and wooly break that
you would rupture a certain number of tubes. And it was
that second parameter, which was essentially treated as zero
for a large number of tubes in the risk assessment, because
if we didn't have any knowledge of a way to get a larger
number of tubes ruptured than a few.
MR. CATTON: I didn't know you -- give it a
number.
MR. LONG: You said give it a number?
MR. CATTON: Well, now you have a way to get that
large number of tubes.
MR. LONG: Well, we have a hypothesis.
MR. CATTON: I don't know if it's real, but--
MR. LONG: We have a hypothesis, but the trouble
is if you put in a conditional probability of zero, you'll
end up about where we did in 1570 for the other types of
degradation. And if you put in a number - a conditional
probability of one, you'll end up where they prioritization
for the GS-123.
MR. CATTON: EDA. I thought we were doing zero
one.
MR. LONG: Which is -- well, what gets you up to
something like, you know, 3.4 times ten to the minus four
was it? And that becomes a matter of opinion, where you are
in that range if you believe that you can damage a very
large number of tubes, without some way of quantifying the
probability of how many tubes you would damage, you can't
see anything more than you're in that range. But first,
you'd like to know that it's really possible to, with, I
guess fatigue, grow these cracks and damage the tubes.
MR. SIEBER: But none of that has anything to do
with 95-05, right? Nothing in the free span?
MR. LONG: Assuming the support plates stay in
place, then that shouldn't--
MR. SIEBER: Right.
MR. LONG: 95-05 should not be affected by that
phenomenon.
MR. BONACA: But, you know, typically, I mean,
when you don't know, it's not really zero or certainty. You
tend to do sensitivity analysis to make -- get an
understanding of what it could be. Again, I thought I had
read them in the INEL report. They're right there. And so
I was trying to myself personally calculate what they could
be -- to see what -- how reasonable this could be. And a
big dependency actually was operator action.
MR. LONG: Right.
MR. BONACA: Because ultimately you come back with
pretty low with pretty low numbers anyway, if you believe
that the operator can handle it, even if you assume
conditional probability of tube failure to be one. And so,
I mean, I don't think it was that speculative. I just -- I
wanted to explain how -- I mean, I was looking for an
evaluation that would answer the DPO, and I just couldn't
see -- there was a window there that I -- didn't seem to be
covered.
MR. LONG: Okay. Well, I don't think it is
covered if you say that there may be a very high conditional
probability of failing 10, 15, 20 tubes, because the human
error probability in that case pretty much becomes one. So
you really don't have the ability to--
MR. BONACA: No, no, no. We just heard this
morning that there is significant probability of success
after about ten tubes.
MR. LONG: Well, that's what I said. If you go
10, 15, 20 tubes, if you believe that that's possible, with
a significant probability, conditional probability, you'd
have to get that conditional probability down to where it
and the initiating frequency for the wild and wooly steam
line break, just those two together, were low enough to not
matter to your conclusions.
MR. BONACA: I agree in the range. Yes, I agree
with you. If you go above the 10, 15, you really -- it
depends very much on the conditional probability. I agree.
MR. LONG: So the conditional probability of
rupturing, let's say, between 10 and 20 tubes, if it's below
ten to the minus four, you're fine. You don't need the
humans to do anything to keep the net contribution of risk
low.
We're sort of getting over into my next slides.
Trying to put the question you asked about other things that
might be initiated by something other than tube rupture, I
believe you mean, and lead to tube rupture. There are the
secondary depressurizations we've been talking about.
There's also the primary overpressurization events, and just
-- I think maybe I should try to go through these slides
fairly quickly, because we've kind of talked about them.
The potential initiators for secondary
depressurizations are things like stuck main steam safety
valves. We've had a steam dump control problem in the one
plant that doesn't have MSIVs. That was a coning that
resulted in blowing down the generators. Spontaneous breaks
in the main steam line safety valve headers. We've seen
those occur during hot functional testing, pretty
operational. It turned out to be a design problem that they
really weren't designed for reactive loads. So that sort of
brings up the question if they're not designed with reactive
loads with steam, but that's been fixed, now if we talk
about overfill events, and you start putting the weight of
water and the reactive loads of discharging saturated
fluids, now do you have a problem with the breaking the
header, as opposed to just sticking a valve. We have right
now a licensing action from Catawba requesting that they not
have to deal with certain single failures on overfill, and
we've asked them, are you confident that if you overfill
and, you know, discharge saturated water that you are
structurally capable of withstanding the loads. And they
don't know. So they have a conditional failure probability
of point one for those events, and they're -- it's a
risked-informed application. So we're pursuing that. So
there's a variety of ways you could get into having - not
only an open secondary, but maybe an open secondary you
can't recover.
CHAIRMAN POWERS: They put point one on the
conditional probability of an overfill event?
MR. LONG: Sticking a safety. In other words,
they were calculating conditional probability of overfill,
and they were looking for what would take them to core
damage. And their conditional probability of sticking a
safety valve, given that they're discharging saturated fluid
through it was point one. And we were, and, of course, they
have a potential for somehow recovering it if it sticks. So
there's questions of realism, of, you know, maybe you break
the header and you wouldn't have a chance to recover, and
there's also the question, can they really gag a safety
valve while something's passing through it.
The conditional probability of the tube rupture
depends on the probability that there's a susceptible flaw
in the free span and the generator that's affected by the
blow down. And that's something that's part of the DPO.
There's the question of how well can you detect the flaws
there. We've heard some of the POD discussions, but most of
the detection is done with a bobbin coil.
CHAIRMAN POWERS: Maybe you've just been simple
here, but it be in the free span or in the U.
MR. LONG: That's true. And in the U, I don't
think -- Jack, in the U they need to test with something
other than a bobbin coil -- to be--
MR. SIEBER: RPC.
MR. LONG: So it has to be an RPC up there. The
-- I'm losing my train of thought. I guess Calvert Cliffs
had a problem with detecting things in the free span and
actually did a rotating pancake coil, actually a plus point
inspection of a large quantity of the free span, and they
found a lot of things, they didn't find with the bobbin.
They then had to go back and find -- do the same thing again
after a short period to show that those things had been
there for quite a while, as opposed to they were suddenly
growing in rapidly after initiation. But it looks as though
the bobbin coil has the kind of PODs that we were describing
to you yesterday. So there is a potential for missing some
things. It's just a matter for probability.
MR. HIGGINS: And the reason here, again, that you
limit it to the free span is because you're assuming that
the TSP will restrain any cracks that are there?
MR. LONG: For this case, we were assuming that
the type of degradation allowed by 95-05 would not
participate in the ruptures, yeah.
Human error probabilities are real important here.
We've already discussed that; that depending on how much you
rupture it, it may be possible or not possible for the
humans to really respond in a timely way. But even when you
have something that's well within their capabilities, just
like with the spontaneous rupture initiating the event,
there is opportunity for errors of omission or commission to
take you to core damage if you've ruptured the tube. And
the difficulty here is you really have to get to cold
shutdown to terminate this event. Whereas, if the rupture
is spontaneous, you have the option -- the opportunity at
least of getting down to below the main steam safety valve
set points, and if they've closed, you basically have
terminated your LOCA. So this -- it's a little more
difficult.
We were assuming that mitigation of about ten full
ruptures is possible, but we didn't try to -- we didn't have
a frequency for that many ruptures, and we didn't really
push hard on the human error probability there. These
numbers are the things you've already dug out of the INEL
report. But they didn't -- at ten tubes, they really didn't
bear in the risk assessment results at all.
And as we've discussed before, we're kind of
sensitive to the idea that we're looking for mechanisms that
could fail a lot of tubes, and if you bring me one, I'll
certainly put it in the risk assessment. But at this point,
it's hard to put something in that you can't really credit
physically.
CHAIRMAN POWERS: How about blow down forces?
MR. LONG: Well, that's why I'm saying. If that
turns out to be something that looks like it has the
potential, we'll definitely put it in the risk assessment.
To try to be complete, let's talk a little about
actions initiated by overpressure events. The initiator is
really ATLAS.
MR. CATTON: Before you leave?
MR. LONG: Sure.
MR. CATTON: For tubes without flaws, what
probability of failure do you assume to the overheating by
the hot gases?
MR. LONG: Right now, the way the calculations
have been done since NUREG 1150, they're calculated on the
temperature of the surge line and the creep carrier damage
accumulation in the surge line.
MR. CATTON: So you're assuming.
MR. LONG: Versus the one number for the inlet
temperature of the hot tube bundle from RELAP. If you do
that, you get about 20 minutes, if I believe, time period
between--
MR. CATTON: Well, that's not the question.
MR. LONG: So I would get zero on that basis.
MR. CATTON: Zero?
MR. LONG: Zero. Now, when I put flaw in--
MR. CATTON: Well, that's nonsense.
MR. HOLAHAN: Be clear. You're not assuming zero.
You're doing the calculation and in the model, you're
calculating the clean tube temperature and its likelihood of
failure. And there is a model for likelihood of failure of
tubes with no cracks, which I think is what the question
was.
MR. CATTON: That was the question and the answer
is that the probability of failure of the intact, undamaged
tube is zero.
MR. HOLAHAN: No.
MR. LONG: The way NUREG 1150, 4550 did it--
MR. CATTON: I'm not going to -- well, what did
you do in 1570. I know what they did in 1150.
MR. LONG: Okay, I was going to tell you what we
did at that point was essentially the same thing. At that
point in time, we were basically trying to extrapolate from
1150.
MR. CATTON: Oh, okay. Okay.
MR. LONG: 1150 had brought up.
MR. CATTON: No, I understand. I understand.
MR. LONG: Okay, since that time, when I try to do
Farley, I try to take into account something about the
distribution of temperature in the tubes, and I guess, we'll
get into this later, because Charlie is going to talk about
how we do the modeling of the tube temperatures. But the
distribution of the temperature -- RELAP does not really
attempt to calculate the average temperature in the bundle
and the hottest temperature in the bundle. MAP does make an
attempt at that. But they do it with an average temperature
and then they make a guess in plume assumption.
MR. CATTON: In either case, either MAP or RELAP,
they're based on inadequate information. So, I'm just
curious as to what you do about that, when you go into your
world of risk.
MR. LONG: Okay. It probably would be better to
ask me this when we get to talking about what I do for
Farley, because I did try to capture that when I did Farley,
and it would help if Charlie had a chance to do his
presentation first.
MR. CATTON: I don't want to bore everybody else
here, so I can wait.
MR. LONG: I recognize the problem. I was
afflicted with this problem a year ago, when I was really
hard put to figure out what to do with it. So I'd be glad
to explore it as soon as we get everything on the table.
ATLAS, fairly quickly, the ATLAS is a fairly gross
model. It assumes that if you exceed the level C service
pressure for the reactor coolant system that something
terrible will happen and you will damage the core. We
looked at ATLAS events in the Surrey model to try to figure
out if they were part of the high dry. It looked like most
of them weren't, although if you had a failure of all aux
feed, they could be. We had some thermohydraulic cases run.
Len Ward ran some for us where we actually sequentially
ruptured tubes when we reached certain pressures. And lo
and behold, it lowers the peak pressure in the ATLAS. They
would have to be fairly weak, tubes, because in the ATLAS
situation, you probably have the main steam system up near
the safety valve set points so it's a thousand plus PSI.
The primary system is, if it only goes to 3,200 PSI, you're
maybe at 2,200 pressure differential across the tubes.
That's not the full main steam line break pressure
differential. Now, the ATLAS pressures aren't limited to
3,200 PSI. They may go higher. What we assumed was the
potential for getting core damage if you went above 3,200,
and the potential for rupturing the tubes. And if you get
up -- in the way we did it in 1570, you'd add five percent
bypass -- five percent of your ATLAS core damage frequency
to the bypass if you were just blowing the tubes from
pressure effect alone. Since you, 3,200 PSI is a little bit
below what was giving us five percent conditional rupture
probability. We weren't quite sure where we were in the
average rupture probability for all ATLAS sequences. But it
looked like, given the frequency of the ATLAS sequence being
low enough it wasn't really contributing much to our answer.
CHAIRMAN POWERS: I'm thinking about overpressure
events -- you looked at accidents that initiated
overpressure events. I wonder have you thought at all about
an event that involved the relocation of fuel in the water
producing a shock wave?
MR. LONG: I've thought about it. We haven't
tried to calculate one yet. The -- that's one of the things
that gets you way out in the accident, so that if -- what
you'd have to do to get to where you're talking about is to
somehow have gotten the RCS out to where you have major
relocation into a pool of water on the lower head, and not
have a very large hole in the RCS that would, you know,
pretty much allow that wave to--
CHAIRMAN POWERS: As an ardent baysian, of course,
you see this as an extraordinarily likely sequence?
MR. LONG: I'm sorry. Say this again?
CHAIRMAN POWERS: As an ardent baysian, you see
this as a fairly likely sequence, right?
MR. LONG: I'm not a baysian. I hate to tell you.
CHAIRMAN POWERS: Just PRA, and he's not a
baysian.
MR. LONG: No, I get nervous when I see people say
we haven't had a steam generator tube rupture yet. So our
probability is lower than those other guys. We see those in
our submittals.
CHAIRMAN POWERS: Well, you have had a core melt
accident in which you relocated fuel and or water fuel, or
plenum with no -- with the system pressurized?
MR. LONG: With the system pressurized?
CHAIRMAN POWERS: With no effect on the steam
generator tubes?
MR. LONG: I guess the point is--
CHAIRMAN POWERS: No exploding, either.
MR. HOLAHAN: And when the system is at pressure
it's probably less likely to have such a--
MR. LONG: So people claim.
CHAIRMAN POWERS: It's a -- one to push that
database very hard.
MR. LONG: Let me tell you how far we've gotten in
the thought process on this. Frankly, we don't think our
models are very reliable out that far. But to try to keep
the RCS at high pressure, you know, up around the 2,200 or
whatever safety valve set points that far into the accident,
you're really saying that you haven't creep failed anything
first. And it looks to us like you probably would. So we--
CHAIRMAN POWERS: I mean, I've got a -- I've got
one accident, which I melted fuel and PWR, and it didn't
creep rupture anything. Well, it did a couple of spiders up
in above the fuel.
MR. LONG: Okay, it also didn't heat up the steam
generator tubes.
CHAIRMAN POWERS: That's also true.
MR. LONG: Right. Anyway, it apparently relocated
some fuel into water. I understand TMI has had a hard time
being simulated, and it wasn't very cooperative in being
able to be simulated by RELAP. But to try to tell you as
far as--
CHAIRMAN POWERS: But we could put out a generic
letter -- one must only have severe accidents that are
easily simulated by RELAP then.
MR. LONG: But seriously, we did -- we have had
what probably amounts to more or less to a bull session
about this. We've tried to think about it. And this is as
far as we got; that we thought that if we really had the RCS
at fairly high pressures that what would happen would be we
would creep fail something before we, you know, relocated a
lot into a, you know, a pool of water in the lower head. We
thought if we had depressurized substantially, hopefully
there would be some hole. If you pressurize, I understand
the probability of getting a steam explosion is higher,
right.
CHAIRMAN POWERS: Yeah, there is -- I mean, what
-- the conventional wisdom is that triggering steam
explosions becomes increasingly difficult with increasing
pressure to the point that the trigger is equivalent to the
yield, so--
MR. LONG: Right.
MR. HOLAHAN: Some say.
MR. LONG: If the majority voted to that extent,
yeah.
CHAIRMAN POWERS: The -- I mean, the database is
based on some droplet tests, triggering tests. And there --
and people smile about those and say, okay, I think I
understand this, why it might be. Except there's this
obnoxious Winthrop experiment where they pressurized and it
damaged -- the resulting explosion damaged their facility
and they had to quit doing things. So it's a mixed bag, and
I understand that some of the experiments that they had done
in recent past in, I guess, Germany or Europe anyway, that
they too began to question this pressure inhibits triggering
concept. It's not -- the problem is that we just don't do a
lot of steel and aluminum tests in high pressure systems,
where the vast majority of our database on steam explosions
come from. So--
MR. CATTON: That's the history of the steam
explosion, isn't it?
CHAIRMAN POWERS: Oh, yeah.
MR. CATTON: You develop convention wisdom, then
you blow it up.
CHAIRMAN POWERS: That's right. Yeah, I mean,
that's -- I mean, that's certainly the history of the copper
industry and the aluminum industry that they get some idea
of what prevents these things. They pursue that idea until
the next explosion and then they host another conference and
sponsor more research.
MR. LONG: Well, to try to tell you where we got
to -- we were thinking about what would happen if you had
deck tubes and a big enough hole in the RCS to have
successfully depressurized it, and then you drop the
relocation into the pool and created some sort of pressure
pulse. It's not clear to us exactly what the temperature of
the tubes would be at that point, because when you've lost
the density in the RCS, even if you've heated the tubes up,
they probably cool down some just from transfer of heat to
the rest of the structure. But as long as the secondary
side was somewhat intact, it doesn't look like you would
rupture the tubes and raise the pressure in the steam
generator high enough to open safeties. So, the other part
of it is you should then go back to something that looks
like containment pressure. So even if you fail the tubes,
it's not clear you create a very substantial release to the
environment from that failure of the tubes at that time.
Now, it's really -- we haven't thought about it
beyond there. I -- we're having enough trouble with the
things we are trying to think about very hard is the best
thing I can tell you.
Let's see.
CHAIRMAN POWERS: Well, I'm encouraged that you're
thinking about this thing before you gain a great deal of
solace in saying that I want to creep rupture myself out of
this -- out of this problem is to do remember that TMI
didn't creep rupture anything. And we didn't heat up the
steam generator tubes, either. But is there something in
between those two?
MR. LONG: We have tried to ask ourselves some of
those questions. TMI was sort of an intermediate pressure
LOCA. It wasn't sitting on the safety valve set point,
because it was stuck open. Yet, it wasn't down to where the
accumulators would come in, either. And in the license
application that Arkansas submitted last March, they tried
to simulate this by just lowering the safety valve set point
to 1,400 PSI and running that for a bunch of cases. Well,
the difficulty is they did that earlier in the, you know,
the transient, so they created all their loop seals,
saturated at 1,400 PSI, and they kept it there.
When we, instead, put -- started sticking safety
valves open later in the transient and depressurizing
continuously until something evaporated and created more
pressure, it got to be quite more interesting, and I guess
we can show you some slides, if not if you need to. But it
turned out when you opened the hole and how big the hole
was, even though we just restricted ourselves to holes in
the top of the pressurizer, it would still give you some
forced flow past the surge line. It still gave you a
protracted accident, and some clearing and reforming of loop
seals so you were getting -- as some of that water was
evaporating and getting to hot metal, you would get pressure
pulses and things did not look real well behaved. The best
thing I can is that there's a whole very poorly charted
territory there that we don't think we can really give you
the answers for yet.
You asked a question about risk metrics, and I was
assuming this was -- should we use delta LERF or go to human
health effects from the releases. So I wanted to give you a
few thoughts on that. If that was not the question, you
should correct me soon.
We -- we're not really sure what the definition of
LERF is because it seems to change. But -- so it's not
really clear if steam generator tube failures leading to
core damage by various paths do exactly or do not exactly
meet the definition. In doing the licensing work, we've
tried to say, well, if it doesn't quite meet the definition,
but it's close to it, it's sure a lot closer than continuing
to accident source term. We're going to treat it as LERF.
So pretty much anything that looks like the secondary site
is open when the core is being damaged and the tubes are
pretty much from primary to secondary, we're going to treat
as LERF, for licensing purposes.
If we try to go to the full level three
consequence calculations, we still have some problems
getting there from what we know today. We really haven't
fully evaluated the effects of the RCS blow down through the
fault at steam generator, and the -- right now, the tube
temperatures are calculated as if there's no net flow out of
the generator to the secondary, so we have a mixing that's
assumed from the 1/7th scale test that's in the inlet
plenum, and that keeps the temperature down to the tubes;
that we talked about this, I think, on Wednesday a little
bit that if you have some substantial flow out of the tubes,
you are no longer forcing fluid back into the inner plenum
from the cold side. You're sucking it out of both sides,
and the mixing will probably go away. The tube temperatures
will probably go up quite a bit, and it's not really clear
what you're doing to additional failures of the tubes.
We've talked a lot about jet cutting, and we think if it's a
little leak, probably we're not in a jet cutting regime.
It's still not quite clear what happens if the tube that
you're -- is about to melt that you're squirting fluid on
across the way.
So, we really haven't defined the size of the hole
between the primary and the secondary as you progress
through an accident where you're really depressurizing the
RCS into the secondary. So that makes it very difficult to
find the flow rates in the secondary side--what the
velocities are going to be, what the temperatures will
become, what the deposition rates would be for the nuclides
that are transported through there. So it's very hard to
define a source term that is really applicable to these
accidents once you've decided that the secondary is really
becoming opened in a gross way to the primary. And for that
reason, we don't think we're really ready to try to go to a
level three until we can get to some of these, you know,
physical phenomena better at hand, if we ever do. The other
part of it is, if we did go to level three, it's not really
clear what the acceptance criteria are for the consequences.
Do we have a safety goal policy statement that has numerical
objectives for close-in populations, for one-mile for prompt
fatalities, and ten miles for, you know, latent effects like
cancer. But the bulk of the health effects may occur beyond
ten miles, especially if these things are late enough to
allow for evacuation, and many of them would be. So,
there's an issue of comparing what to what.
CHAIRMAN POWERS: I guess the reason -- real
question that we were asking here is there anything about a
basis coming out of the steam generator, secondary side,
especially large releases with bigger things -- a
substantial amount of material out there. It would change
our general view that LERF is a good surrogate for the
safety goal policy statement.
MR. LONG: You say is there anything unique about
them? I mean, they're different from what you would have
from a crack in the containment or openings around
containment penetration bellows or something of that sort in
the sense that you have smaller volumes with more structure
to be transited. I'm not an expert in that area, and I
don't see the person that I would ask that ask that question
here. I don't know what to say about the difference in
terms of the transported radioactive material. In terms of
timing, you can calculate when you think the releases would
start to occur, and depending on the size of the leak from
the primary to the secondary, it may be quite late in the
process, so there may be something like a not early large
release that would be a better comparison. And I know for
the boilers, there's an issue of late failure of containment
that is also sort of in this category.
CHAIRMAN POWERS: They -- I mean they have a
long-term station blackout. It's kind of funny beast to
deal with. It seems to me that the real concern is that
they could be very early in the accident sequence.
MR. LONG: These releases?
CHAIRMAN POWERS: Yes.
MR. LONG: Certainly a fast station blackout, you
know, could be pretty early. And if you could get to a very
large -- you know, LOCA outside containment due to something
like the wild and wooly main steam line break with the
massive tube ruptures or leaks that could be fairly early,
too, especially if you had any failures in ejection
processes. Right now, we model them as if everything works,
and you've got a -- you know, flow out the RWST. So,
there's a wide range. I know a lot of the IPEs originally
came in with core damage sequences not counted because they
resulted in core damage after 24 hours for spontaneous leaks
-- spontaneous ruptures, I mean. So, in that regard, it's
late from evacuation standpoint, but it may still be early
from the standpoint of time for settling radionuclides in
the system. So in that regard--
MR. HIGGINS: Steve, this is a -- maybe you
haven't done this, but maybe get your opinion. If you took
the end of site -- a typical end of cycle leakage estimate
from 9505.
MR. LONG: Okay.
MR. HIGGINS: And you ran a one of these risk
metrics on it. A delta LERF. Which region of break I.1.174
would you fall into in evaluating that change?
MR. LONG: Okay, let me answer part of that first,
because you said region, that brings me into a couple of
different parameters at the same time. Research did run a
case where they assumed a 100 GPM leak from primary to
secondary at the time that essentially has started. And
they ran it all the way through with melt core, including
the containment. And they allowed the failure in the
containment by the surge line failure to, you know, occur on
the model as opposed to keeping it from occurring and seeing
how it long the tubes to fail. So what you really do is
once you breach the RCS pressure boundary in the
containment, you drop the driving force of the -- you know,
pushing the radioactivity out the hole in the steam
generator tubes. And that drops the dose to the public
quite a bit. So Charlie's going to have to see if his
memory is better than mine maybe if he gets up here, but it
seemed to me for 100 GPM, primary to secondary leak rate
size hole, assuming that hole does not become any larger
during the accident, and the secondary was open, Charlie, we
ended up with something like a factor over what was assumed
to be a contained accident. Is that right? We can -- and
this was assuming more than the -- 1150 assumed more than
tech spec value of leakage from the containment to the
environment. So that's also a little bit of a shaky
baseline.
But it did not look like it got you into the LERF
range, if that was the size hole, and you knew that having
that size hole did not alter the accident so that the
failure was still into the containment, and the reason is
that you're not very far into the core damage phase of the
accident before you relieve the pressure on the -- you know,
the tube and stop driving so much through the tube wall.
Now, when you're asking where does that put me in
-- Reg Guide 1.174 regions--
MR. HIGGINS: Yeah, whether or not you cross over
into above a ten to the minus sixth change in LERF or--
MR. LONG: What I'm saying is it wouldn't be a
LERF, so you'd be doing it on core damage frequency.
So I wouldn't think that you would, and I'd have
to first ask what am I starting with from core damage
frequency.
MR. HIGGINS: Okay, that's good enough.
MR. LONG: Okay.
MR. STROSNIDER: This is Jack Strosnider. I would
like to add one observation there, and I think you know --
you did -- it's probably a reasonable question to say what
region would you be in in Reg Guide 1.174 to talk about the
delta. That means you have to understand what the
probability of tube rupture was before the generic letter
was implemented. We don't have a good handle on that, but I
think, you know, it's -- it wasn't assessed specifically,
but as I indicated yesterday, if you look at what people
were doing before generic letter 95.05, before the voltage
based criteria, they were attempting to size these defects.
And we talked yesterday about, you know, the ability of NDE
to size defects. And, of course, this was back before some
of the methods that are available today were available. So
I would just suggest, and this is just my opinion that the
probability of leakage from one of those tubes prior to
95.05 wasn't zero. Alright. So, that's the delta you'd
have to assess. And we don't have a, you know, quantitative
answer to that, but I think you need to consider where we
started and where we went to.
MR. LONG: If it's not zero, it's pretty close.
MR. KRESS: Yeah, but the real delta I think to
not be interested in is the thermally induced failure of the
steam generator tubes so that it becomes a large leak and
what's the probability of that compared to the probability
of pulling the search lights.
MR. LONG: I agree, and I think that 95.05 has no
effect on those cases, because they're not the vulnerable
parts.
MR. KRESS: I think I agree with you. It doesn't
matter whether you had bad tubes or good tubes, it will go
about the same time, I think.
MR. LONG: Well, I'm not prepared to say that.
What I am prepared to say if something goes in the steam
generator, I don't think it's the very short cracks
underneath the tube support plates that were allowed to stay
from 95.05.
MR. KRESS: I hear you. That's why -- that's why
I say it doesn't matter whether it's good tubes or bad
tubes. They both go about the same time.
MR. STROSNIDER: In the for what it's worth
department, when I presented, when we had the discussion
with CRGR on generic letter 95.05, I suggested that, in
fact, 95.05 could be improving the situation versus what
people were doing in the past. And I still -- I still think
that. It got a little bit of debate, but nonetheless I
think that sort of puts it in perspective.
MR. LONG: Well, it certainly initiated things
like, you know, condition monitoring and -- you know,
operational assessment processes, that I think have been a
big help.
I'm a little bit ahead of the agenda here by going
into the risk metrics before some of the other subjects I
have on, so at this point, I think probably I want to put up
the thermal hydraulic calculational part, if Charlie's
ready.
MR. TINKLER: I'm Charles Tinker, from the Office
of Research. The objective of my presentation is to briefly
review the severe accident analysis of--
CHAIRMAN POWERS: Turn things, the red light comes
on. Dead battery, again?
Sam, check the switch on the bottom.
MR. TINKLER: Oh, there we go. That was it.
Gentlemen, I'm going to have to bring my reading glasses to
the -- again, I'm Charles Tinkler from the Office of
Research. The objective of my presentation this afternoon
is to briefly review the severe accident analyses, and its
underlying bases that was used to evaluate the
thermohydraulic boundary conditions that might be seen by
steam generator tubes during a severe accident. And the
focus is on thermally induced failures of steam generator
tubes.
MR. KRESS: With what purpose in mind, Charlie?
MR. TINKLER: Well, the reason we e did these
calculations was in support of NUREG 1570 to look at things
like conditional failure probability of tubes during some of
these kinds of accidents. And actually, I kind of
remembered our numbers of condition tube failure
probabilities, but they were in the context of flaw
distributions, typical average severe flaw distributions in
plants.
MR. KRESS: Yeah, the reason I asked the question
though is are you looking to see if there's a significant
risk associated with this that we have forgotten to analyze
before so that it might be worthy of looking at a back fit
or something like that?
MR. TINKLER: Well, I think the idea was to look
at incremental risk from changes in the steam generator tube
criteria.
MR. KRESS: These are 95.05 incremental risks?
MR. TINKLER: No, I don't think it was -- I don't
think it was in connection with 95.05, but we looked at it,
for example, on electrosleeves.
MR. KRESS: Electrosleeves. Yeah, I remember
that.
MR. TINKLER: Whether was there -- was there any
incremental risk by adopting the electrosleeve repair
process. Did we increase the probability of a thermally
induced tube rupture, and my own sense was that in the NUREG
1150, there wasn't as much focus on the sequences that
involved the secondary side depressurization, which is yet
another failure and makes the overall sequence probability
smaller, but there wasn't quite as much attention as we're
devoting now to those sequences that involve the additional
failure of the secondary site to remain intact and at
pressure. Because that has a two-fold effect, and I'll talk
about this more. It obviously increases the delta p across
the tube, but it also increases the thermal load on the
tubes, because the reduced pressure on the secondary side
provides a smaller heat sink in terms of the steam on the
secondary side, so you -- in our calculations, we can
increase the temperature of the steam generator tubes by on
the order of 100 degrees K -- between the pressurized
secondary side and a depressurized secondary side. And that
makes a fair amount of difference in terms of the thermal
loading on the tubes.
Along the way, I hope to address a number of
issues that have been -- that have been raised for a number
of years now, and some of which are repeated in the DPO.
I don't -- I'm going to skip -- I have lots of
viewgraphs, so I'm going to skip through some of them. You
can see the outline. We've talked -- we know what the issue
is. To point, too, that natural circulation and transfer of
heat through the loops of an RCS was identified some number
of years ago--generally, it was thought to be a good thing.
Gets heat away from the core. Distributes it through the
system. It allows for the fortuitous depressurization of
the system to prevent bad things like high pressure melt
ejection and DCH and things like that.
But if you depressurize the secondary side, by
having a secondary side safety stick open, then you do
produce a challenge to the steam generator tubes.
And this is the cartoon that we normally show to
represent the natural circulation paths. I'm going to
deviate a little bit just because has been raised a number
of times. But the question often comes up, we produce all
these calculations that show counter current natural
circulation and creep rupture. How come it didn't happen to
TMI? Briefly, there are a few key factors that influence
natural circulation. First and foremost is the pressure in
the system. Higher pressure systems produce greater natural
circulation. Higher density flow to convect heat away from
the core. It also produces greater density differences, so
it's two-fold effect.
The RCS configuration. The U-tube configuration
steam generators are, by their nature, more likely to draw
flow than the once-through steam generators. It is very
difficult to get steam to go down through a once-through
stream generator and back up after it's dried out. It
doesn't happen. The tests at the University of Maryland
show that you can't get natural circulation so that big heat
sink out there, isn't there. So you have nothing to draw
flow away from the core. So you produce a weaker natural
circulation pattern. They do see natural circulation in the
hot lake, but it's a reduced effect compared to this.
Core blockage. If you form blockages in the core
region with crossed around them, there's no way to get from
inside that material out into the loops. And if you can
intermittently inject water at various times during the
transient, and float up over the core, like turning on the
2-B pump at TMI, you shut off natural circulation. Goes --
there's no hot core. You've covered it with water.
MR. KRESS: So, it's not surprising TMI.
MR. TINKLER: Well, TMI, if you look, they had
only a few periods when they could have gotten natural
circulation. And this shows -- this initial -- this is the
initial core heat up. This is turning on the 2-B pump, and
you can see the water level is rising back up during those--
MR. CATTON: They never did serve the loop seal,
did they?
MR. TINKLER: Well, when they turned it on -- they
might have cleared it briefly, but it refilled quickly.
Because if you got a loop seal, you can't get it. You have
to -- well, you can still get counter current, but counter
current--
MR. CATTON: Where's it going to go, to the top of
the candy cane and back?
MR. TINKLER: Yeah, that's all it's going to do.
MR. CATTON: That's all it's going to do.
MR. TINKLER: And also, at TMI, the pressure's
low. They had a PRV that was leaking.
MR. CATTON: The U over tube is too small to get
any recirculation within it. So--
MR. KRESS: With the candy cane, I'd be very.
MR. CATTON: You're just dead in the water.
MR. TINKLER: You can get a little bit in the
candy cane. But it's not a vigorous natural circulation,
and during that first large period where natural circulation
could have occurred, the pressure in the system is low. No,
this is a RELAP calculated pressure, but I do -- I -- we're
pretty good on -- you know, everybody's pretty good on
pressure. But RELAP, it made a loop. We've had a long
time. But we do this calculation pretty good.
And if you look at TMI during this initial period
here, this initial period, the pressure in the system is
quite low, and it's generally acknowledged that once you get
below about eight MPA, it's hard to get a lot of natural
circulation and convect heat away. So I know I was asked
that question quite some time ago, and I generally refer to
deviations from the typical severe accident, okay. And
there were deviations from the typical severe accident, like
reflooding, but I might have neglected to mention that there
-- that the fundamental design doesn't lend itself as much
to that.
But it causes me to think that maybe we ought to
look at some of those typical TMI calculations to try to
focus on how much natural circulation we could have gotten
and see if we can match some temperatures a little better in
parts of the system.
Also, there's an A&O calculation -- some A&O
calculations that were recently done, and these show the
effect of system pressure. One's sitting at relatively high
pressure safe -- at the safety. And one's with a leaking
PRRV. And this shows just the hot leg temperature. So over
this -- in this initial period here, the effect of pressure
makes a pretty big difference.
It's not meant to be an exhaustive treatment of
it, but it at least provide a little more clarification,
because I would agree that if you've only had one accident
to look at, and it didn't produce the thing you say happens
all the time, it could cause you to wonder.
MR. KRESS: Well, Tim, I probably run the risk
dominant sequence.
MR. TINKLER: I don't want to address that.
CHAIRMAN POWERS: Before you go to this
inter-circulation through the steam generator, I'd like to
understand a little better about the free loop circulation.
MR. TINKLER: Okay.
CHAIRMAN POWERS: When Steve was talking earlier,
I got the impression of an increased interest in this and
that it introduced an enormous amount of complexity into
this situation.
MR. TINKLER: Well, the loop seal clearing is --
it's a -- you know, first you -- you got to do more than
clear the loop seal. You also got to get the water level
below here. Okay, the down comer skirt. Because if all you
do is clear this, but you don't clear this path.
MR. KRESS: That's another loop seal.
MR. TINKLER: That's another loop seal. Right
here. So, but now we're able to clear both of them in a
number of calculations as a result of boil off and just
general water dropping in the core. And when that happens
you produce full loop circulation. And the key is what's
going on in here, because when you produce full loop
circulation, you don't get any cold flow returning through
the steam generator to dilute what goes into the tubes.
Another way to look at that is turning to page 17. This
shows temperatures around the loop at the time that we
normally predict surge line failure for Surrey. And if you
look at the temperature coming in from the hot leg, the
1,500 degrees K, the reason we don't instantaneously fail a
lot of tubes real quickly is because it's being mixed with
cold flow returning back through the steam generator tube
bundle. Okay, it's being mixed and diluted. And the reason
it's being mixed and diluted is because we have a loop seal.
If we didn't have a loop seal, it wouldn't be quite this
high, because there would be other things going on. But
we'd have a whole lot higher temperature passing through the
steam generator.
So when we do calculations where we produce loop
seal clearing, the issue is whether or not the pressure in
the RCS has dropped low enough at the time a loop seal
clearing occurs, because if it hasn't dropped a lot, we
predict failure of pristine, intact unflow tubes.
MR. KRESS: How good are you at predicting when
the loop seal clears?
MR. TINKLER: Well, we think we can predict loop
seal clearing reasonably well. It's a question--
MR. KRESS: When you have three loops?
MR. TINKLER: Well, it's the question of whether
or not we can predict which loop seal clears.
MR. KRESS: Yeah, that's the question I was really
asking.
MR. TINKLER: And we don't believe that we have
enough confidence in our prediction of which loop seal
clears, so when we approach this probabilistically in 1570,
we assumed an equal probability among the loops. We didn't
-- because we calculated loop seal clearing in some cases,
and we typically calculated in the loop where the safeties
haven't stuck open on the secondary side.
MR. KRESS: And if you're in a loop that doesn't
have the surge line?
MR. TINKLER: Right, it was a loop that didn't
have the surge line, and it was loop where the secondaries
didn't stick open.
MR. KRESS: Yeah.
MR. TINKLER: And if you're looking at a loop
where the secondaries didn't stick open, these sequences
where loop seal clearing typically involve some
depressurization of the RCS, because you're boiling water
out of the loop. That's what clears it, and in those
sequences we actually had a higher pressure on the secondary
side than on the primary side. Okay. So we wouldn't have
predicted failure. But for 1570, we ignored that. We
assigned an equal probability to clearing these loops, even
though we always predicted it to occur in a loop that --
where the secondary side was not depressurized. So--
MR. KRESS: But if the secondary side is
depressurized, and even if you were in the leg where the
surge line was, it -- I was under the impression that you
were -- your calculations would almost there at times show
that you failed the steam generator before the surge line
under those conditions.
MR. TINKLER: If it's a loop where the secondary
side is not depressurized, no.
MR. KRESS: That's not true if it's not
depressurized.
MR. TINKLER: That's not true, because typically
these sequences with loop seal clearing involve some
depressurization of the RCS, of the primary side, so you're
-- those will be loops where the secondary side will be at
1,000 and the primary side will be at 600 or 800. So we
can't buckle these tubes, you know. We predict they're hot,
but they won't fail.
MR. KRESS: A different failure mechanism in that
direction.
MR. TINKLER: Because the pressure is the other way
now on those cases. But when we did it, when we looked at
for assessing conditional tube failure probabilities, we
ignored the fact that we were actually predicting it in the
other loops and assigned a uniform probability to its
occurrence, because there is considerable uncertainty as to
when you predict loop seal clearing and in what loop you
predict it to occur.
MR. KRESS: That's what I thought.
MR. TINKLER: That is true. And it was a dominant
-- it was a dominant contributor to -- I believe -- tube
failure probability. It was the big deal.
MR. KRESS: That's what I was -- I was under the
impression of, too.
MR. TINKLER: That is correct.
CHAIRMAN POWERS: When you say the pressure is --
gets with the secondary sides still pressurized, and the
pressure in the primary is now below the pressure in the
secondary, when does accumulator dump occur? And when it
does, do you then Jack the pressure back up?
MR. TINKLER: Well, in some sequences where we had
-- where we were modeling reactor coolant pump seal leakage,
you would see some cases where, when we got down to
accumulator set point, for example, we'd get some flow being
driven into the steam generators, and that would cause them
to, in some cases, cause those tubes to heat up fairly
substantially. But typically speaking, reactor coolant pump
seal leakage and leakage in general or the RCS, unless it
produced loop seal clearing generally didn't cause a
problem. If it produced loop seal clearing, then it did,
because we gave it equal probability. But there is a nuance
associated with depressurization, where you get accumulator
injection and then you force steam flow into the steam
generator, okay, without the benefit of a lot of mixing,
because then you're -- then you have almost -- you know, in
those cases, you force it through both paths of the hot leg.
So those cases did produce some more, but it's a relatively
short-lived transient where that occurs.
CHAIRMAN POWERS: I was just wondering if your
tubes were hot, and you got a dump so that it jacked the
pressure so that you had the delta p across, you'd just get
a rupture, even though it was a short transient. Well, you
know typically we don't show radical pressurizations on
accumulator injection. We get a little pressurization and
then accumulator injection stops. We had an issue where we
looked at this where we were condensing additional water in
the cold leg, and that was causing us to eject more from the
accumulators. And that's an issue we've had some
discussions with the industry folks, because they contend
that we inject a little too much water as a result of that.
Because they show a very smooth accumulator pressure
injection transients. Those are a little more spiky, a
little more ragged. So, but -- that is a nuance that has
come up in some of the calculations.
MR. CATTON: I don't quite understand your figure.
The 1504, 982, and 775, what are they?
MR. TINKLER: Well, these are -- these are
calculations of intermediate volumes in the inlet plenum,
okay. These -- this is, in effect, a mixture temperature.
MR. CATTON: So do you feed some of the tubes 1504s
and some tubes 9--
MR. TINKLER: No. No, these two streams--
MR. CATTON: Are mixed?
MR. TINKLER: Are mixed according to the mixing
fraction.
MR. CATTON: Which is?
MR. TINKLER: Point nine. So 90 percent of the flow
is at this temperature, and twice as much of it is at this
temperature.
MR. CATTON: How do you get the 982? Where does it
come from?
MR. TINKLER: The 982?
MR. CATTON: That's again a mixture.
MR. TINKLER: That's the result of 90 percent of
this flow being mixed with this 775, okay. See, this cold
flow returning through the steam generator bundle.
MR. CATTON: Sounds really complicated. Where did
you get the information to base that on?
MR. TINKLER: Well, this is -- these values were
deduced from the 1/7th scale, in effect, deduced from the
1/7th scale test data.
MR. KRESS: Yeah, I'm interested in how you
actually made that deduction. Did you have temperatures in
the -- certain tubes of the steam generator?
MR. TINKLER: Well, they had rotating rake thermal
couples in the inlet plenum.
MR. KRESS: Okay.
MR. TINKLER: And they have temperatures in the --
about an inch or two in the tubes, up in the tubes, in the
tube sheet.
MR. CATTON: In some of the tubes.
MR. TINKLER: In some of the tubes.
MR. KRESS: Did you have a temperature in the hot
leg?
MR. TINKLER: Oh, yes. There's temperatures
throughout the system. You know, in the hot leg -- in the
top and bottom of the hot leg.
MR. KRESS: And did you have a way to deduce the
flow rates in--
MR. TINKLER: Yes.
MR. KRESS: In the two counter current directions?
MR. TINKLER: Yes.
MR. CATTON: The flow rate was deduced by an energy
balance. It was not pressure.
MR. TINKLER: But they can do a little better job
up in the tube volume.
MR. KRESS: I was going to use the flow rate at an
energy balance to get the mixing fraction is what I wanted
to do.
MR. CATTON: You can't do that.
MR. KRESS: You can't do that that way.
MR. CATTON: Because it was the energy balance that
gave the flow rate.
MR. TINKLER: And, in part, the mixing fractions.
But there's also the observation of mixing from the thermal
couple data itself.
MR. CATTON: Well, yeah, but you see two people can
disagree, and we disagree. There was a meeting held at
Argonne, which I attended, where we discussed all these
things, and the people who were there was Viscanta, myself,
Ishi -- was Griffith there? You were there.
MR. TINKLER: Yes.
MR. CATTON: Peter Griffith.
MR. TINKLER: Peter Griffith.
MR. CATTON: And the way -- the conclusion by
Viscanta and myself, Griffith was kind of neutral, was that
you couldn't really scale this data. There were just too
many unknown factors. You couldn't scale it to the full
size. This was the conclusion of those people. Ishi felt
you could scale it, but his background is two-phase flow.
It's not natural convection, and this is the buoyancy driven
problem. And in the inlet plenum, it's a highly complex,
multi-dimensional flow. When I looked at the temperatures,
I could find a tube or two where the temperature was very
high, much higher than in any of the other temperatures. It
was almost as if it fingered through directly into the tube.
So these kinds of things never became a part of this
problem.
Well, what does all this mean? First, if there's
zero mixing, the tubes will surely fail. If you have high
mixing, the surge line will surely fail.
MR. KRESS: Not surely because the time and
temperature were still pretty close together.
MR. CATTON: Even then, they're relatively close
together, and there are a lot of things that I can talk
about the other side, too. The way the surge line is
treated, the heat transfer is probably not high enough.
Because unless you guys have done something different in
RELAP-5, you still used it as filter. And the heat transfer
coefficient to the surge line should be augmented. On the
other hand, there is some surge lines that come in on the
side. And if that's the case, then the surge line is not
going to be heated as fast. The more you move the surge
line down, the more buoyancy and its effect on the heat
transfer changes. When it's up, you get -- it's probably
helpful. If it's down, it's on the other side.
MR. TINKLER: Well, actually having a horizontal--
of this horizontal leg on the surge line does -- can be a
help, too, because it also helps establish natural
circulation.
MR. CATTON: Well, there are a lot of factors.
There's even the interaction between the two flows and here,
the divided into two pipes. What do you do with something
like this. I think you almost have to give it a -- unless
you want to do the kind of basic research that's needed to
address this complicated problem, you're going to have to
give some credibility to the fact that the mixing isn't
going to be what you think it is. Now, I suspect that, you
know, if you had to make a guess, you guess 50-50 chance.
Who knows where it's at? It's somewhere between zero and
one. And it's certainly not either.
CHAIRMAN POWERS: I come back to my baysian
instincts, even if Steve isn't an ardent baysian, I am. And
they got a test here that has some flaws to it. But comes
back indicating relatively high mixing. I don't no whether
it's 90 percent or 87 percent.
MR. CATTON: I didn't come to that conclusion when
I looked at the data. When I looked at the temperatures, I
came to the conclusion that there were some tubes that were
going to be fed almost directly the high temperature gas.
MR. TINKLER: I guess, we -- I'd have to say, we do
not come to that conclusion. And, you know, this Committee
has, I think, been provided with the results of that peer
review, so, you know, you can take a look to see what --
there were a number of discussions. I think it was nearly
unanimous that the tests were well designed and well
executed and that they indicated mixing. Now, we can argue
about whether or not it's 90 percent mixing fraction or 60
percent mixing fraction or things like that, okay, but we
did -- we have done sensitivity studies on these parameters.
I'll talk about them a little more. And you can see the
effect of them. Whether or not a fluid stream line can go
unmixed from the hot leg up into the tube sheet, I guess is
a, you know, is a concern that has been expressed. We don't
deny that at all. The general indication, though, as far as
we're concerned is that the data does not indicate unmixed
flow. Does that mean it couldn't occur under a range of
conditions, including tube leakage. Well, that's something
that needs a little more consideration. But, you know,
that's the general -- that's the general view we have at
this point.
My first--
CHAIRMAN POWERS: Before you proceed, now, this
peer review that you're speaking of was the same meeting as
Ivan was speaking of?
MR. CATTON: That's right.
MR. TINKLER: Yes. Yes.
MR. CATTON: We each, I guess we read the letters
written by the people who attended the meeting differently.
CHAIRMAN POWERS: Yeah, apparently so. I guess you
have to read them yourself to come to that conclusion.
MR. TINKLER: Well, you know, there is some
questions. For example, we can't scale, in a 1/7th scale
test, the exact flow conditions for a tube, because we can't
make the tubes 1/7th diameter. They'd be too doggone small,
and the hydraulic diameter would be too big, and the
resistance through the tube bundle would be huge. So you
got to have the right flow area through the -- this 1/7th
scale tube bundle relative to the flow area in the hot leg.
And you have to have the right mass. Because it's the mass
of steel that's actually the source of natural circulation.
So it's hard for us to claim that we're simulating each and
every tube.
Now, are we producing the same kind of bulk mixing
pattern in the inlet plenum? We think we are. The ACRS
what used to be the severe accident and thermohydraulic
subcommittee -- I'm not sure what it is now -- but we made
presentations where we compared frood numbers in the test to
the frood numbers in our code calculations, showing that we
were doing a pretty good job of predicting them, between the
plant and the experimental facility. But there are
undoubtedly distortions in that facility that cannot fully
accommodate, you know, the exact nature of mixing in the
tubes. But the other point I make from time to time, with
varying degrees of success, is that the fluid stream lines
are not fixed. Fluid that comes from the hot leg in a
single stream line, and you saw the CFD code calculations.
We can calculate stream lines very accurately if we want to,
but that doesn't mean they say; that what comes out of here
always goes to this one tube out of 3,000. It moves around
a little bit, this plume. Actually, they saw evidence in
the test that the tubes carrying hot flow and cold flow
occasionally change a lot. So--
MR. KRESS: But particularly if you're in a
transient.
MR. TINKLER: So if you got a stream line that's a
little hotter than the average, there's no reason to think
it stays in this tube for a particularly long period of
time. That plume does -- there is some oscillation to it.
Now if -- you know, as I say, I make that argument with
varying degrees of success, so--
But it -- the first summary is that we've used the
SCDAP/RELAP code to analyze this for potentially risk
significant scenarios. And typically, we predict the
failure of the hot leg or surge line before unflawed tubes.
We've done a number of sensitivities on thermohydraulic
modeling. It didn't alter the conclusion, but the margins
are pretty small.
I can skip through this example calculation if
we're--
MR. CATTON: What might be -- do you have one that
shows the time?
MR. TINKLER: Well, I can show as part of this --
I'm sorry, Dana, did you?
CHAIRMAN POWERS: Well, go ahead and answer Ivan's
question. But the question I'm going to ask at some point
in the discussion is that suppose we don't fail the surge
line, is there anything about -- if we do not fail the surge
line, you will predict a steam generator tube failure
someplace, at some time.
MR. TINKLER: Well, typically, if we don't fail the
surge line, the next thing that fails is the hot leg.
CHAIRMAN POWERS: Okay, leave them both out.
MR. TINKLER: Leave them both out?
CHAIRMAN POWERS: Yeah, let's just--
MR. TINKLER: Yeah, we'll fail a tube. Yeah.
CHAIRMAN POWERS: Okay, is there anything about
that tube failure that would be worsened or improved by the
peculiarities introduced by generic letter 9505, or is it
such a robust failure that it's like you're full loop seal.
You had failed a pristine tube just as likely or just about
the same time as you would fail one that's got a few cracks
in it?
MR. TINKLER: I will turn to people much more
qualified to comment on the nature of the failure than
myself. Someone in the front row back there, preferably Joe
or Bill, if they could comment on the nature of that
failure.
CHAIRMAN POWERS: Now, we do have.
MR. TINKLER: Typically, what we assume is that
it's a cross section of a tube for the calculation. We have
done some--
MR. CATTON: Can I help you out?
MR. TINKLER: Calculations of fission product
inventory released off site, okay. Not level three per se,
but fractions of our inventory.
MR. CATTON: Isn't 9505 restricted to that big
thick plate on the bottom?
MR. HIGGINS: No.
MR. CATTON: Or even the tube support plate? The
heat transfer to the plate is going to be enough that that's
going to be a cool spot along the tube.
CHAIRMAN POWERS: Okay. I mean, clearly we do have
this peculiarity of the leakage flow that can change this
whole picture here. But I'm going to leave that out, just
like I'm going to leave out all these surge line and nozzle
failures, and ask if there's anything -- what I'm asking is
how much time to devote to thinking about and reviewing all
of these things. If, in fact, there's -- leaving aside the
leakage question, right now, there's nothing, I mean, it
would fail if I had a brand new steam generator in there
with alloy 690 and no cracks, it would fail just as much as
it would with one that was filled with lots of non-through
wall, non leaking cracks.
MR. KRESS: I certainly believe within the
uncertainties of this thermohydraulic analysis, you can't
tell the difference.
MR. SHACK: There's no uncertainties. He's just
killed the hot leg failure and the surge line failures, and
the only thing that's left is whether the core will rupture
or that will happen.
MR. KRESS: No, what he's asking if there were
uncertainties in these things is such that maybe you do at
some probability fail the steam generator tubes first at
some probability because of the uncertainties in everything.
Would you have gotten the same answer whether you had your
tubes or not.
CHAIRMAN POWERS: What I know is that people that
do these calculations--
MR. MUSCARA: For that temperature that -- you
know, on the transient reaches 840 degrees. So a much--
MR. KRESS: And it doesn't matter whether they're
cracks or not.
MR. BALLENGER: I read 1,200. I mean, 1,500K,
1,200C.
CHAIRMAN POWERS: That's the gas temperature.
MR. BALLENGER: That's the gas temperature.
CHAIRMAN POWERS: It's hot stuff.
MR. BALLENGER: It's hot stuff.
CHAIRMAN POWERS: What I know is that people have
tried to develop codes other than the one that was used for
this calculation, and when they tried to model the counter
current flow, they have to do it the same way RELAP does by
putting in these figures, and things like this.
MR. CATTON: It depends on how much money you want
to spend.
CHAIRMAN POWERS: Well, these guys didn't spend--
MR. CATTON: A really good example of that was the
Comik School from Argonne, and the PTS issue. The whole
nuclear industry uses 1020 now because they don't want to
spend the money on the computer time. So somebody hired a
consultant from CHAM in Huntsville and said, gee, how many
would I need to really do it right. He came up with a
number over 100,000. So what do they do, they say, okay, we
don't want to that.
CHAIRMAN POWERS: Okay, well, there--
MR. CATTON: If they're willing to do that, you can
handle counter current flow. The problem is one of how much
you're going to spend on the computer.
CHAIRMAN POWERS: These guys, you know, they're
independent of these, and so they made different decisions,
though inherently the model is about the same. Okay, you
would castigate it just as much as you do this one. And,
but they did it differently, and, as a result, they
presented curves that were just like those except the labels
were permuted. And so I'm saying if I have that case, and I
assume that's reality, is there anything unusual about this
-- these steam generators now that we've allowed generic
letter 9505 -- other than leakage. We'll put that aside,
because we're going to get to that one a little later --
that have changed the positions of those curves, and I get
the strong feeling that to the level of detail that these
calculations are typically done, no.
MR. KRESS: That's what I feel.
MR. BALLENGER: I mean, is there any error. What
are the error bars on these numbers?
MR. TINKLER: Well, we're going to talk a little
more -- we'll get to that a little more.
MR. BALLENGER: I mean, that's -- if it's 200
degrees, and man this is--
MR. KRESS: Yeah, that's a very legitimate
question. That's why I asked him that initial question.
Dana, I asked him that initial question: for what purpose
are you doing this. And that was the reason, because--
CHAIRMAN POWERS: Well, I know the purpose he's
doing it, because we asked him to--
MR. KRESS: Yeah, I know, but, you know, maybe he
has an alternative ulterior motive, but that was my reason
or that question, because if you perceive there's no
difference, what are you going to do with those numbers? Is
it a new set of risk sequences that you just forgot about
before, and you want to see if they're important or not.
CHAIRMAN POWERS: Well, I think the -- I mean, the
issue that is very important is if we allow the leakage, and
we stipulate that we believe that--
MR. KRESS: Yeah, that may be a significant issue.
CHAIRMAN POWERS: Everything they told us about the
mixing and we stipulate that they simulate the Westinghouse
data out to the third significant figure, and there's
nothing wrong with data, and I admit that questions have
been raised about it, but if we stipulate that and then we
introduce this leakage over -- Now that's an interesting
issue.
MR. KRESS: Yeah.
CHAIRMAN POWERS: Then the question then comes
back, again, is there anything different now if you had--
MR. KRESS: And that is certainly different, but I
think what you probably will find out is if you make the
calculation of the risk that you get due to the -- assuming
the steam generator tubes fail first, you're probably can
make an argument of acceptable risk, but that's something
I'm hoping they get to.
MR. STROSNIDER: This is Jack Strosnider. I'm not
sure, I want to enter into this discussion. What I'm sure
of -- but I guess the one thing I would point out is, in my
understanding of the events being talked about here is that
they're not the extreme blow down events. You know, they
don't put those kind of loads on support plates, et cetera.
And we've discussed, to some extent, the pass that with
regard to the ODSEC at the support plates. The support
plates will be there, so it's not clear to me that, you
know, that's the location that's going to be critical in
terms of tube failure. In fact, I think it's probably going
to be someplace else.
MR. TINKLER: Well, actually -- but these
temperatures are the first region above the tube sheet.
MR. KRESS: Yeah, but it doesn't matter. If you
induce the leakage, it doesn't matter where the leakage is.
It's going to suck the -- you know, it's going to induce
some failure somewhere else.
MR. CATTON: It will probably suck it from both
directions.
MR. KRESS: Yeah, but--
MR. CATTON: If it's a big leak.
MR. KRESS: Yeah, it will change things markedly in
terms of its failure, even though you don't -- even though
you think the leakage is going to be, failure doesn't
matter.
MR. SHACK: Since we're firing off speculation
here, I'll go with Jack. I mean, if I put this thing in
that collar, that thing is not going to have any gross
failure. You know, you're going to see one of my hippo type
failures somewhere else in the free span of this thing. But
what you will get with the generic letter I think is some
leakage through the cracks that's, you know, on the order of
ten gallons under a main steam line break, which would
correspond to some equivalent area, which you can presumably
use to get a gas flow at this temperature.
MR. KRESS: Yeah, and the question is, does that
change this picture?
MR. SHACK: And you'll get some -- well, the
question is whether that additional leakage bothers you very
much, yeah.
MR. HOLAHAN: No. My answer is no. Of all the
things we don't know, which you hear a lot of, the effect of
9505 is not one of them. I think we're pretty clear that
9505 is the least important risk implication.
MR. KRESS: So I guess the real question, from a
risk standpoint, is whether you increase that leakage over
and above what you say is in 9505?
MR. HOLAHAN: Right.
MR. KRESS: Because there's some probability of
that being much greater.
MR. HOLAHAN: Right. You recognize.
MR. KRESS: I think that's a question that will
change the risk--
MR. HOLAHAN: Right. Approving 9505 allows
effectively leakages from going from one GPM to potentially
a little more than that. Okay. And in a realistic point of
view, I think maybe it wouldn't change it all. But at least,
to say, you know, in theory, a virtual leakage call it,
okay, we would allow some.
I think it has no effect on the likelihood or
consequences of tube ruptures or multiple tube ruptures for
any of these sequences. Zero. Minimal. Negligible. Zero.
MR. KRESS: I think you're probably right.
MR. TINKLER: I just showed this. I always show
this so I can overlay this other plot and show you that,
indeed, it's the hydrogen generation --- the onset of
hydrogen generation that really causes things to heat up
here.
MR. KRESS: Yeah, because that's where all the
energy is.
MR. TINKLER: That's where all the energy--
MR. HOLAHAN: To be fair, I didn't get to see the
overlay.
MR. TINKLER: Well, let me, actually, I always show
it on an expanded plot, too, because, you know, depending on
what part of the transient you look at it -- if you look
from time zero, well, it's a small fraction of the time of
the total transient, but if you look at when things really
start to happen, the time differential between tube failure
and surge line failure is a larger fraction of that
interval. Another way of looking at the margin is, if you
look at the time surge line failure is predicted to occur,
and look at the temperature of the tubes at that point,
that's 950 -- about 957. Now, in this calculation, we
predict the tubes to fail at about 1150, okay. Now, Joe, I
just checked with him, he said when he ran his tests, the
tubes failed about 1,110K. Alright, we got to stay on K
here. So 1,110 to 950, that's another indication of the
margin. We're actually--
MR. KRESS: Or it's an indication of the level of
uncertainty.
MR. TINKLER: Well, but it's -- I mean, you say,
15, 20 minutes, that doesn't sound like a lot of time, but I
don't know. There's 160--
MR. SHACK: At 160 per minute, it's a lot of
temperature.
MR. TINKLER: At 160 -- but 160, you know, 160
degrees sounds may be a little better when you start talking
about the sensitivity studies.
MR. CATTON: But I don't have to change the mixing
very much to get that curve?
MR. TINKLER: Well--
MR. KRESS: And then you divide that.
MR. TINKLER: Overall conclusions. Now they -- I
haven't proven these conclusions from the viewgraphs you've
seen, so--
MR. KRESS: These are speculating--
MR. TINKLER: No, these are valid conclusions. We
just don't have enough time to -- for me to show you all the
calculations. But as I said, this is worth a 100 to 150
degrees Kelvin in the tube temperatures, typically. I think
that's about -- in the neighborhood. So that's just worth
about 1,000 PSI across the tubes. So those two factors
combined make these kind of assumptions the most dominant
assumptions in the calculation. If the operator is able to
open the PORV, this problem goes away. We've done quite a
few calculations that show you depressurize -- you can
generally get down to about two and a half megapascals, and
that's enough for this problem to go away, if you can find a
way to reliably do it.
Pump seal leakage. It's biggest effect was on the
loop seal clearing, but it may have some effects on other
calculations, but they appear to be of less importance than
the pump seal leakage.
CHAIRMAN POWERS: Now, pump seal leakage is
becoming less of a problem for plants now, because they put
the improved?
MR. TINKLER: Well, we did the calculations with
the new -- with the distributions for the new pump seals.
But -- it's still a pretty high rate depending on, you know,
the calculations, but with -- with -- indisputably, certain
thermohydraulic boundary conditions and phenomenological
issues are important in the plenum mixing. It's clear, if
you don't mix at all in the inlet plenum, that makes a big
effect on your calculation. We think there is inlet plenum
mixing. Heat transfer modeling makes a difference, and loop
seal clearing makes a difference.
CHAIRMAN POWERS: Can I ask you a phenomenological
question? If it takes too long to answer it, tell me so,
because it may not be germane here. As you have that
counter current flow going along the pipe leading into the
plenum, that's modeled as a fairly smooth process. It's not
really. And it won't be very long smooth. Does that
disrupt any of this -- any of these arguments or any of
these thermohydraulic modeling?
MR. TINKLER: I'm not sure I understand your
question.
MR. CATTON: The interface between the hot stream
and the cold stream will be both friction and heat transfer,
and this will reduce the impact on the steam generator
tubes, and as far as I know, when we did work on it, we
didn't include it. And they certainly don't by sticking it
in pipes.
MR. TINKLER: No, we don't, but the observation
from the test data is that those streams are fairly
isolated, and there is not much mixing between the streams.
That was the--
MR. CATTON: It depends on the velocities.
MR. TINKLER: It does, but I can only tell you that
the general conclusion from that test data was that it is,
those streams are fairly well isolated. Now we did
calculations to model heat exchange between the two streams.
That makes it better. That's good for us. It lowers that
average temperature going into the steam generator, getting
a little more mixing, a little more heat transfer between
those two streams, lowers our peak tube temperature. That's
to the good, and we do calculate -- we did calculations that
maybe I'll get to you that will show you -- that will at
least show the numbers.
MR. CATTON: But it's kind of like Los Angeles,
Dana. You know, if yo fly in there, you can see the top of
the smog layers just as smooth flat surface. And it's
diffusion controlled. So whatever you do, because the hot's
above and the cold's down below, you transfer it from one to
the other, it's going to be -- I mean, everybody's flown
into Los Angeles.
CHAIRMAN POWERS: Just to be indulgent, since I'm
the chairman, I get to do these things.
[Laughter.]
You know, you got aerosols in the hot stuff that
want to go down. And they go down pretty good rate.
MR. CATTON: That's okay. But that's a little bit
different. I mean, I -- that's still a bit different.
MR. TINKLER: That's -- you know, I had thought
about it. But you know, they actually did some tests in the
1/7th scale to look at the effect of aerosol deposition in
the hot leg. They primarily looked at it from the
standpoint -- they didn't actually -- I take that back --
they didn't model aerosol deposition, they put a heat source
on the pipe to see if that disrupted the natural
circulation.
CHAIRMAN POWERS: Oh, and that's a good piece of
information.
MR. KRESS: I suspect you were also using steam,
and you got through telling us that this temperature really
took off when it was the hydrogen generation part, and I
don't know how hydrogen would behave under those conditions,
either.
MR. TINKLER: Well, we -- you know, we have
hydrogen in our calculations, and they did inject a simulant
for hydrogen in the 1/7th scale tests.
MR. KRESS: Oh, I didn't know that.
MR. TINKLER: Yeah, they had, there were five
separate phases to their -- you know, their high pressure
test program. I think one of them included a lighter gas
than sulfurhexaflourine. So--
MR. HOLAHAN: I mean, there's absolutely a minimal
amount of racinium.
CHAIRMAN POWERS: A minimal amount of what?
MR. HOLAHAN: Of racinium. Just thought I would--
CHAIRMAN POWERS: Oh.
MR. HOLAHAN: Throw that in.
MR. KRESS: That's what I was expecting, yes.
CHAIRMAN POWERS: It's an all steam system, so that
we wouldn't expect it--
MR. TINKLER: Actually, I thought you were
prompting that -- I don't know whether there was something
later on in this sequence that might make this sequences a
little different. The chimney effect if you fail something
and later fail the vessel.
Code validation. We've talked about this, so I
won't dwell on it, but, you know, we didn't -- we didn't
just start doing these kinds of calculations. We've been
doing them a long time. And the folks at INEL, Len Ward and
Darryl Knudsen, who's here in the audience today, whose done
more of these calculations than anybody in the world,
probably everybody else in the world combined, actually.
We've done a lot of them.
CHAIRMAN POWERS: I think we're willing to
stipulate that.
MR. TINKLER: Okay.
CHAIRMAN POWERS: I wonder if we could -- just a
few to -- schedule a little bit, take a recess at this
point, come back and discuss this effect -- the section of
effective leakage on inlet plenum mixing.
I mean, I don't want to take out things that you
think it's important for us to hear.
MR. TINKLER: No. No.
CHAIRMAN POWERS: But on the validation of the
model and the basis for it, I think -- we're willing to
stipulate.
MR. TINKLER: Sure. Okay.
CHAIRMAN POWERS: Those things and then move to the
issue that's part of our contention, which is the effect of
leakage on the mixing. With your indulgence, and I
appreciate that, we will return at a quarter after and
resume on this section.
[Recess.]
CHAIRMAN POWERS: Let's come back into session. I
apologize for interrupting your presentation, Charlie.
MR. TINKLER: Okay.
CHAIRMAN POWERS: And, again, if there is material
that I suggest we jump over and you think it is critical, --
MR. TINKLER: Well, I would just like to very
briefly talk about the sensitivity studies --
CHAIRMAN POWERS: Sure.
MR. TINKLER: -- that show what the effect of some
of these parameters that are of debate.
We talked about the parameters that influence
mixing and the temperature in the tubes. Some of the
parameters identified early on were the number of tubes
carrying hot flow, the mixing fraction, the recirculation
ratio.
We went back and looked at the range of values
deduced from the 1-7 scale test data and varied those
parameters for the calculation. Single sensitivities varied
over the range showed a change in the tube temperature on
the order of 20 degrees or less, so they didn't seem to have
a large effect.
DR. KRESS: Those are kind of weird looking ranges
to me, .76 to .89. How did you arrive at what to choose for
those? 29 percent, why not 30 or --
MR. TINKLER: Well, we took the numbers that were
evaluated from the 1-7 scale test without rounding them up
or down, or --
DR. CATTON: So there is no consideration of the
possibility that --
MR. TINKLER: They could be different.
DR. CATTON: Rare probability that there was some
error in the scale.
MR. TINKLER: We will address that later, and I
will talk about that a little later. We did some additional
calculations in response to recommendations made by the ACRS
and by the peer reviewers. They said, well, that range of
parameters you changed was pretty narrow, some of the
comments you just heard, and why don't you change a couple
of things at the same time?
So, first, we changed heat transfer coefficients.
And, generally, whenever we changed heat transfer
coefficients, it made things better, because --
DR. CATTON: It depends which one you change.
MR. TINKLER: Well, it depends which one you
change. But, remember, this is our base case. We only had
one where it went the other way, all the other temperatures
got lower. And that is because the environment in the steam
generators is nearly adiabatic. The tube, the difference
between the vapor temperature and the tube temperature is
really quite small. So we can't change a heat transfer
coefficient and make the tubes hotter. We can make the
other stuff hotter but we can't make the tubes hotter.
MR. BALLINGER: What you are saying is is that
this calculation is not -- is dominated by something other
than what you varied?
MR. TINKLER: Yes.
DR. CATTON: It is dominated by the mixing.
MR. TINKLER: Yes.
MR. BALLINGER: Completely.
MR. TINKLER: Yes. Although, if we increased the
heat transfer coefficient at entrances more than 1.3,
because you could argue that maybe that is not enough for an
entrance effect in some local geometries maybe, we could
maybe improve the performance of the tubes relative to the
hotleg or surge line.
MR. BALLINGER: But how far off could the dominant
thing be? What does dominate?
MR. TINKLER: Well, I mean if you think there is a
probability of unmixed flow going to the steam generator
tubes, you go back to that 15 --
MR. BALLINGER: So is there a real estimate of the
uncertainty?
MR. TINKLER: Not yet.
MR. BALLINGER: Like we have going around this.
MR. TINKLER: Not yet.
MR. BALLINGER: An uncertainty on that.
MR. TINKLER: Not yet. We will get to that. We
did a simultaneous change of parameters using the 5 percent
confidence limits from the test, the transient test, which
we believe to be the most relevant test for these particular
calculations. And when we changed everything, assumed they
were all independent and changed them in the worst
direction, to the 5 percent confidence limits, we increased
the tube temperature 50 degrees.
But that is still a mixing fraction of .73, so it
is not like it is -- it is not like we changed the mixing
fraction to zero.
DR. CATTON: Or even to 50 percent.
MR. TINKLER: Effect of leakage on steam generator
inlet plenum. Concern has been raised that steam generator
tube leakage during severe accidents could alter mixing in
the inlet plenum. The 1-7 scale test did not simulate tube
leakage. The idea is basically that -- and I had that out
there for so long. Well, the argument is that you have
3,000 tubes drawing from the inlet plenum, or, actually,
roughly 1500 tubes draw hot flow from the inlet plenum, and
maybe one of them now is drawing a lot more flow than all
the other tubes. So is it going to disturb that mixing
pattern in the inlet plenum?
At first observation, these tube leakage effects
may very likely be disbursed among many tubes. It is an
aggregate sort of thing, it is not one tube, and if it is
disbursed over the tube bundle, you would be hard-pressed
that it is going to dramatically influence it.
Leak area equivalent to a 1 GPM leak is a very
small fraction of the tube bundle flow and the inlet plenum
flow, the flow circulating in the inlet plenum.
CHAIRMAN POWERS: The numbers we discuss in
connection with predictions from one cycle to the next and
whatnot are all much higher than one gallon per minute.
MR. TINKLER: At 100 GPM, it is about 10 percent
of the inlet plenum flow. Now, is 10 percent spread out
over many tubes likely to influence the inlet plenum mixing?
DR. KRESS: Is 10 percent of one tube likely to
influence it?
MR. TINKLER: Well, I am not even sure that it
makes it worse, frankly. I mean drawing more from one
location may have the influence of, you know, people use
jets to mix things.
DR. CATTON: But you also have a buoyant plume
down there somewhere, and you might just suck away the fluid
that is mixing and then becomes the hot fluid.
MR. BALLINGER: You are not firing a jet into
something, you are sucking something out.
MR. TINKLER: Yes, I know, I got a jet coming out.
It is an exit jet, as opposed to -- but, so the bulk
velocities in the inlet plenum are not likely to be
influenced a great deal. The velocities at the inlet to
that tube, if it was one tube, would be quite higher, much
higher. So if the mixing occurs down deep in the inlet
plenum, then you might not expect the effect to be large,
but if the mixing occurs up close to the tube, you know, to
the tube sheet, it could be a more significant effect.
DR. CATTON: These are buoyancy driven processes,
and there was an experiment by Myinger some time ago where
it actually had to do with core melt, but just a small
fractional variation in the density, he put this bubble into
a mixture, and it just wipes everything out. You don't need
to do very much to completely disturb whatever the pattern
is that is there.
MR. TINKLER: Well, the general issue of mixing
and tube to tube variations is more problematic for any
codes like this.
DR. CATTON: You are absolutely right.
MR. TINKLER: So what we have laid out in response
to the user need received earlier this year from NRR is a
plan to look at this specific issue using the more detailed
CFD codes. We have in-house expertise that has been applied
to CFD codes, developed over several years, and we think we
can take a look at this to at least provide insights as to
the magnitude of the influence of this tube leakage. Does
it radically alter the mixing patterns?
And we think it is promising, we think it will
allow you to look at other things, too, other sensitivities,
the location of the entrance of the hotleg and things like
that on the inlet plenum mixing.
DR. CATTON: Do you make any distinction in the
calculations as a result of location of the surge line?
MR. TINKLER: Surge line?
DR. CATTON: Yes.
MR. TINKLER: Yes. Yes. We distinguish between
surge lines that are oriented with a horizontal leg or, you
know, an initial horizontal and vertical, or just a vertical
riser, yeah, we do.
DR. CATTON: So do you know where the interface
between the hot and the cold is? I guess -- no, I am not
sure you do unless you have a velocity.
MR. TINKLER: Well, I was referring to the
orientation of the hotleg at the inlet plenum steam
generator. But this, we would use this to study
specifically the issue of inlet plenum mixing, the general
issue of inlet plenum mixing and to gain insights as to the
effect of tube leakage on that mixing.
But for small leakage rates, it is clear it is a
small fraction. At 10 percent of the inlet plenum flow
rate, it may not be very clear that you will be able to
distinguish much difference either, especially if it was an
aggregate leakage over many tubes. It would be very
difficult to draw a conclusion about that. But if it is
isolated, perhaps much more so.
But, in any event, we do believe this will -- this
is, you know, these are the kinds of codes that were
developed for these kinds of issues, so we think it is a
good application.
CHAIRMAN POWERS: My experience with CFD codes is,
in truth, zero. But my witnessing of those calculations is
that the CFD codes do a very fine job if you have some
experimental data to compare against. And the kinds of
experimental data they compare against usually are
substantially more detailed than what I think you have
available on this mixing in the 1-7 scale test.
Have you given thought to the feasibility of doing
the experimental investigations that would be useful for
comparison of the CFD code analyses?
MR. TINKLER: We have. We have thought about
commissioning experiments to look at this specific issue.
One could conduct perhaps simpler experiments to look at
plume mixing in more idealized configurations, as opposed
to, you know, steam generators. That was just pretty
complicated at some level.
But the very first step in doing this will be the
validation benchmarking of the code against available data.
DR. CATTON: Which means Westinghouse, right?
MR. TINKLER: Well, which includes the
Westinghouse 1-7 scale test data. If I came back here, or
if Chris Boyd comes back here and tells you about his CFD
calculations, you know, a year or so from now, and he
doesn't compare them to the 1-7 scale test data --
DR. KRESS: We would wonder why.
MR. TINKLER: You would want to know why. So,
now, that doesn't say that that is fully dispositive on it,
so we are looking at that now, and we are in the first
stages of undertaking that particular activity.
DR. KRESS: How are we supposed to factor that
into our --
MR. TINKLER: Well, I think that --
DR. CATTON: You can't.
DR. KRESS: I know, I mean --
MR. TINKLER: Well, it depends on the leakage rate
you want to consider. If you want to consider --
DR. KRESS: I want to consider the leakage rate at
least that you have in 95-05, that it allows.
MR. TINKLER: Up to 10?
CHAIRMAN POWERS: Up to 130.
DR. KRESS: 130, 150, something like that.
MR. TINKLER: 130.
CHAIRMAN POWERS: They tell me they get very
nervous when they go to 130. Try 130.
MR. TINKLER: Well, like I say, 100 GPM is about
10 percent of the flow rate. That is not an overwhelming
fraction of that flow in the tube bundle or in the inlet
plenum. So, --
DR. KRESS: But it is getting up there where you
might think it could have an effect.
MR. TINKLER: It could. I guess I would be
tempted to say it would be a greater effect if it were a
point source as opposed to a spread over some large number
of tubes.
DR. KRESS: Oh, sure. Sure.
MR. TINKLER: Okay.
CHAIRMAN POWERS: Yeah. But I am not sure how
spread it is, because certainly they showed us an example of
a tube with -- my recollection is that one cycle it was on
the order of seven gallons per minute, and on the --
projecting it forward to the next cycle, some higher number.
So I am not sure how spread it is.
And on top of that, from what I see in these
patterns of steam generator repairs and whatnot is that the
most highly damaged tubes seem to come in clusters. They
may not be spread over the entire diameter.
MR. TINKLER: Well, you know, I guess I would be
tempted to say it would be -- it would be nice to have
experimental data upon which to draw some simulant fluid
test to look at mixing plumes with a -- while you are
drawing a jet off perhaps in an isolated region. That would
be a nice supplement to the calculations, because we will,
in effect, be extrapolating.
But the code, you know, I think that the code will
have the capability to look at this issue in a reasonable
way. But I don't know what else that I could tell the
committee at this point.
DR. KRESS: It looks like very difficult
experiments to do because geometry is so important.
MR. TINKLER: It is, it is.
DR. KRESS: You almost have to do a full scale on
those.
MR. TINKLER: Well, I would just be concerned
about preserving the general, you know, aspect ratios and
flow areas.
CHAIRMAN POWERS: But it seems to me --
DR. CATTON: If it were just the natural
circulation within the plenum region, it is just one
parameter, geometric similarity, and you can scale the Relay
number or the Grashoff number. But the fact that you feed
it some amount of flow, you have probably got a Reynolds
number in there, too.
Water is probably the thing to use, because you
can get a very high Relay number and it is probably going to
be turbulent and that is going to give the CFD codes a
headache because they still haven't really got there with
good turbulence models. And when it is buoyancy driven, you
have to treat all of the Reynolds stress terms. And it is
doable with CFD, there is no question. But I am not sure
that if you pick up a commercial CFD program, you are going
to get all that you need. You have a very nice paper on
this.
MR. TINKLER: The good news is it's single-phase.
CHAIRMAN POWERS: And it seems to me that in
wrestling with the experimental issues, which I think are
formidable, based just on the critiques that have been
labeled on the 1-7 scale, the overall strategy seems to me
like a pretty good one to start with the calculations and
calculate the bit, small, and in between, and things like
that, and at least get a feel for what's doable.
I think he has a real challenge in getting this
geometrical similitude here.
DR. KRESS: I do, too. I think there's a real
challenge there.
MR. TINKLER: I don't know how much -- we're
running a little behind.
DR. CATTON: The problem is that if you use a
simulated fluid, and you want to get a high number, you're
going to go to a liquid. As soon as you go to a liquid, the
final number gets big, and that there, the number is less
than one, or in gases, it's at most an order of one.
DR. KRESS: You can't simulate all of that.
DR. CATTON: That creates differences in the
mixing process, but it's on the conservative side.
MR. TINKLER: We are undertaking some new work to
further resolve some of these issues of uncertainty which
heretofore have been addressed through sensitivity studies,
and combinations of sensitivity studies.
We're going to look at different accident sequence
variations. An awful lot of calculations have been done on
-- a lot of sensitivities have been done on the Surrey
plant, and we're going to look more at a Zion type design.
We will, indeed, be looking at independent mixing
and tube-to-tube variations. SCDAP/RELAP will be used as
the principal tool for the system level analysis, okay?
But we will be using the CFD codes to look at
things like in the plenum mixing, and also tube-to-tube
variations, because the CFD code provides the kind of
resolution to look at those kinds of things in greater
detail.
DR. KRESS: Let me ask you about this new
research, and use an eight-letter. Does it have anything to
do with the DPO issue?
MR. TINKLER: I don't think so. I think the
calculations that were done for NUREG 1570, 15-20 minutes.
There's a kind of a sense that's, you know, that's not a lot
of time.
Things go differently than what you think, and a
good 15 to 20 minutes becomes minus five minutes, so --
And I think there's a sense that as we do more and
more of the assessment of delta risk and risk impacts, that
we need to look at the uncertainty in some of these
calculations, especially where the margins appear to be a
little small, and look at them more rigorously.
So, like I say, what we want to do is develop
distributions for these parameters, and they may go outside
the range of values seen in the experimental data, you know.
Like all distributions, we'll have tails on
distributions, and we'll argue about what those tails on the
distributions will be, and we will peer-review this, okay?
So, we'll have a couple of more opportunities to
do discuss what constitutes mixing and a characterization of
mixing.
These are the parameters we've initially settled
on, but we'll consider that also, I think, as part of the
peer review.
MR. HOLAHAN: I'd like to answer the question
about the relevance to the DPO. I think the answer is that
it's not related to the DPO issues.
If anything, this sort of analysis provides you
insights as to what is really important, and I think it
reinforces the fact that issues like 95-05 are not dominant
sequences.
DR. KRESS: Thank you.
MR. TINKLER: Actually, this is just a repeat of
the things I said just a couple of second ago about --
MR. HIGGINS: Excuse me. You said they were not
dominant sequences, but this hasn't been done yet, so what
do you base that on?
MR. HOLAHAN: I base it on that I don't see any
relevance to what this has shown or will show to the failure
of short axial cracks underneath the support plates.
MR. LONG: This is Steve Long to add a little to
this. In terms of relevance to a DPO, the user need was not
-- help us with the DPO; the user need was written primarily
because we developed a large number of issues that we were
having difficulty with to try to move this into
risk-informed regulation.
I will get into some of the applications when I
talk about some of the problems in the next slides.
MR. TINKLER: The Committee asked to hear a little
bit about fission product deposition, the issue of
deposition of fission products on the tubes, and that
contribution to heating of the tubes, specifically in
relationship to the work that was done and published by
JAERI.
These are points that I discussed a number of
years ago in presentations before the ACRS, but basically we
used the Victoria Code to calculate the fission product
release, transport, and deposition.
The Victoria Code is specifically a fission
product chemistry code with provisions for modeling
transport and deposition, but the thermal hydraulic boundary
conditions, pressures, temperatures, flow rates, are all
provided to it by the SCDAP/RELAP calculation.
And basically what you see here is that the
volatile fission product release is on the order of ten
percent decay heat.
That's a fairly consistent number that you will
see in a number of these calculations, at least insofar as
the early phase of core melt is involved.
We predicted that the fission products were spread
among the upper plenum, hot leg, steam generator plenum, and
tubes.
I won't dwell on that, unless there are questions.
Similarly, I'll skip over the Victoria nodalization.
CHAIRMAN POWERS: It seems to me that there was
one line that is pertinent from that slide on the Victoria
capabilities that came up yesterday. Maybe you weren't
here.
The question was raised on whether you treated --
you definitely were here.
MR. TINKLER: Yes.
CHAIRMAN POWERS: Treated a agglomeration and
thermophoresis --
MR. TINKLER: Yes, we do. We treat that, along
with laminar deposition, terminate deposition, settling, and
that's pipe bends, not pipe blends, okay?
And we can talk about some of the additional
models that maybe one needs to consider when they model
fission product deposition on the secondary side, as you're
concerned about the release, but that's not the issue for
this, but we do, indeed, model thermophoresis.
CHAIRMAN POWERS: One of the questions that has
emerged in recent years on thermophoresis is a question over
whose model is best. And my understanding is, without a
great deal of knowledge in this subject, is that the SOFARIS
code being developed by the Europeans uses a different
thermophoretic model than the Victoria Code.
MR. TINKLER: Well, I'm not sufficiently familiar
with SOFARIS thermophoresis models. I know that discussions
of differences in thermophoretic deposition have occurred as
a result of comparisons between some of our calculations on
FEBUS and some of the European calculations.
Frankly, we see oftentimes the prediction of the
thermal hydraulic boundary condition as being more important
to that comparison than the details of the thermophoresis
model, because we often end up with greater differences in
the prediction of the difference between the vapor
temperature and the wall temperature, okay, especially when
you're trying to predict deposition along a thermal gradient
tube where there is relatively steep gradients.
But, again, our steam generator tubes and the
vapor are about ten degrees apart. It's quite difficult to
imagine that thermophoresis --
DR. KRESS: Your use of the term, laminar
deposition, is probably going to overwhelm it.
CHAIRMAN POWERS: That, of course, raises another
important thing. I think we have to bear in mind that -- I
think there are two things:
I think that it is true that these calculations
don't have thermophoresis as a dominant deposition mechanism
throughout the length.
And the other is that theoretically, we don't have
a validated way of simultaneously depositing things by
multiple mechanisms.
DR. KRESS: That's exactly that each of them are
assumed to be independent, and I don't know really how you
-- you have to -- to get thermophoresis, you have to convert
your bulk mean temperature difference that you calculate
with something like SCDAP/RELAP into a temperature gradient
near the wall, actually.
And I'm not sure how you do that in Victoria. I
don't know what you're inputting.
CHAIRMAN POWERS: I think that I do know how they
do that. I think they have a fully developed correlation
and they just match them.
DR. KRESS: They just match each, and then they
get a laminar layer, and that's the distance they get for
the delta-T, okay.
CHAIRMAN POWERS: I think it's built into the code
to do that.
DR. KRESS: Okay, you just put the heat transfer
coefficient into the input.
CHAIRMAN POWERS: I think they just use a fully
developed flow correlation.
DR. KRESS: They recalculate it themselves.
CHAIRMAN POWERS: Yes, they keep track of it as a
function of the flow velocity. I do know it's fully
developed flow. I mean, that's about all I know about it,
and that raises all kinds of questions about whether you
should be doing fully developed flow in these things.
I just thought it was useful to make sure that
that went on the discussion record here, because the
question was raised yesterday.
MR. TINKLER: Yes, well, again, we don't see it as
a dominant mechanism in virtually any parts of this
calculation. So, we think there's an explanation as to why
it was cited as a dominant mechanism by others.
Okay, I'll get to that briefly.
DR. KRESS: The Japanese cited it as a dominant
mechanism?
MR. TINKLER: They cited it as a dominant
mechanism.
Why don't I just go right to that? They used
SCDAP/RELAP also to drive their code, which is ART, not to
be confused with ARTIST, but aerosol release and transport,
who knows. It could be.
CHAIRMAN POWERS: It really doesn't sound
Japanese, does it?
MR. TINKLER: No, it doesn't. They also conclude
that the surge line failed first, but they had a rather
substantial fission product heating of the tubes.
And the main reason is, they assumed that the
temperature of steam entering the tubes not quite equalled
-- this may be a little bit of an overstatement -- it wasn't
quite equal to the temperature of the hot leg, but it was a
lot hotter than ours.
DR. KRESS: It didn't have the mixing in there.
MR. TINKLER: They had a temperature difference of
250 degrees. They just assumed.
And the best we can figure, after numerous
discussions and e-mail and conversation, is that they wanted
to conservatively estimate deposition due to possible
thermophoretic effects.
I guess it's also true that the entrance
temperature to the tube bundle isn't readily apparent from
the SCDAP output. We don't have that intermediate volume,
so, you know, if you're looking at SCDAP output, you've got
a choice of these things.
Well, we don't give you -- the output doesn't
automatically include that mix that's then also compensated
for by the mixing fraction and the ratio of the flows.
So, using a higher temperature, using a
temperature that's 15 times higher than ours produces more
thermophoresis. But, frankly, we just can't see any way
whatsoever you could get that kind of temperature difference
between the vapor and the tubes where the secondary side is
depressurized and there's no water.
Now, the people running the experimental facility
in Europe, the ARTIST facility, they're contemplating
looking at large thermophoretic deposition rates, but that
might be associated with putting some water back in the
steam generator where you can create large temperature
differences, in which case you could get that.
But the other issue -- you know, the obvious is,
if we the temperature that much hotter going into our tubes,
will it fail because the steam's too hot? I don't care what
the thermophoresis is.
The other point is if you think this is an
entrance effect, then it is in the tube sheet and I don't
know, maybe I am going out on a limb here but I guess that's
the last region I would worry about a lot due to fission
product heating anyway.
There the dominant mechanism was gravitational
settling at the top because it is a long distance and
actually Jason reminded me that it is liquid through much of
the system, so if you did deposit a little bit at the tube
sheet, it might be liquid. It might drip off and go down
into the inlet plenum and be on the bottom of the steam
generator.
Conclusions -- we have analyzed tube heating
during severe accidents using benchmark models validated
against scaled experimental data. It's undergone peer
reviews. Sensitivities have been examined. We have seen
temperature variations between 20 and 50 degrees.
We have evaluated tube performance during severe
accidents. We think that further evaluation though would
benefit from the resolution of thermal hydraulic
uncertainties. We have plants to undertake that work. We
think that a more rigorous consideration of uncertainties is
warranted. We think there's something to be gained by
looking at additional sequences for different plants and we
think there is a role for more detailed CFD modeling in this
calculation of details related to the mixing issue.
I do have a couple of viewgraphs on offsite
release. You had it in your agenda. It wasn't really a
part of a lot of our work. I didn't know if you were
interested in seeing anything about that or not. It is
basically the Victoria calculations that were done assuming
a tube rupture about the time of -- we simply ignored surge
line hotleg ruptures and modeled the tube rupture and we
continued the calculation until we predicted the hotleg
would have melted, okay?
DR. KRESS: Did you include the secondary building
and --
MR. TINKLER: Not the building.
DR. KRESS: Not the building?
MR. TINKLER: We did include the secondary side of
the steam generator.
DR. KRESS: Secondary side of the generator
itself?
MR. TINKLER: Of the generator itself, but not
additional deposition in the --
DR. KRESS: Once it got out of the secondary
side --
MR. TINKLER: It was out. It was out.
DR. KRESS: And you looked at both the control
room and --
MR. TINKLER: No. No, we were just looking at
fractions released.
DR. KRESS: Oh, fractions released.
MR. TINKLER: Fractions released, yes. These are
fractions of core inventory released.
The reason we didn't have more noble gases
released is because we released them through the PORV.
DR. KRESS: Okay.
MR. TINKLER: But we had about a 30 percent iodine
release -- so -- that's a real iodine spike.
DR. KRESS: How come the cesium gets to be so low
in this? Let me see it again.
MR. TINKLER: Yes. Cesium released from the core
is only 35.
DR. KRESS: I have always wondered about that.
MR. BALLINGER: Cesium is highly soluble, right?
MR. TINKLER: Yes. It's going to be cesium, most
of it in this calculation would be cesium hydroxide.
DR. KRESS: You know, a lot more of it got
retained in the primary-secondary than the iodine. That's
what -- that one is one that bothers me, I guess.
MR. CHAPAROW: This is Jason Chaparow from the
Office of Nuclear Regulatory Research.
The releases from the core, as you can see, are
limited to about three-fifths of the core, if you look at
the nobel gases and the iodine and the cesium is not far
behind it.
In this sequence we had, after the tube rupture we
continued to get accumulator injections and that kept the
lower part of the core down below about 1500 K so the lower
two-fifths we really didn't predict much fission product
release until this hotleg melted and you just -- the rest of
the steam boiled off so the lower area of the core was
predicted to be a little bit cooler, cool enough to prevent
the fission product releases prior to hotleg melting.
That affects all of the releases to the
environment by almost, by 40 percent.
MR. TINKLER: We heat the whole system up by
continuing this calculation. We just get revaporization of
iodine and it goes out --
DR. KRESS: Okay.
MR. TINKLER: -- and we did a brief comparison
against the early, early MAAP calculations on this thing.
For some reasons their release wasn't through the
PORV so they had more of it go out but on the iodine release
it is about the same.
DR. KRESS: But it is a large release?
MR. TINKLER: We consider that a significant
release.
Four hours though may be judged to be --
DR. KRESS: May not be a large early release.
MR. TINKLER: May not be early. Actually the
calculations typically produce, typically would involve some
evaluations, so there's not much going on prompt.
DR. KRESS: So it may not be kosher to equate this
directly with the large early --
MR. TINKLER: No, not if you are talking about
four to nine hours, four to eight hours, something like
that, maybe not. Well --
MR. HOLAHAN: Well --
MR. TINKLER: Well --
MR. HOLAHAN: Well, it's sure not small.
MR. TINKLER: My comment was how early was early?
This surely was -- I didn't say large. We said significant,
but it is, the difference between significant and large in
this case may be small.
[Laughter.]
MR. HOLAHAN: I think there was a comment earlier,
maybe Steve Long made it, and that is when you have the
choice between treating cases like this as large early
release or as core damage with basically no release, they
look more like the large early releases.
DR. KRESS: It would be prudent to do that.
MR. HOLAHAN: It would be prudent to do, yes.
[Pause.]
MR. LONG: I wanted to clean up a couple of
things.
First of all, I made a comment when I was up here
earlier about the amount of radioactive material that would
go out from the hundred GPM size hole and the tubes if you
went through the station blackout core damage accident
sequence to the point where you failed the surge line, and
then went ahead and failed the surge line in the
containment.
I tried to grab the document during a break and
grabbed the wrong document so I think we need to owe you
that document. My memory is probably not good and Charlie's
memory is better about how much of the radioactive material
went out from that particular case and how it would compare
to a contained reactor accident.
I think my memory is probably good that it wasn't
approaching LERF but in terms of the multiples of the
contained reactor accident releases were probably not on
target.
Another thing I would like to do is there was a
question about whether or not the tubes having flaws in them
made a difference when they would fail. I wasn't sure if
that was a question about if the tubes were 9505 tubes
confined in the support plates or if they were free span
flaws, so if they are free span flaws it will definitely
make a difference and it depends on what is going on in the
RCS in terms of heatup and pressure changes.
This isn't probably the best slide that I should
have. It is just a slide that I happen to have. What we
did is we modified RELAP/SCDAP to take account of tube
temperatures with different stress multipliers so it
simulated tubes with different size flaws instead of looking
just at the pristine tube.
I think you can kind of read it from your chairs
but the black is the weakest tube and it is something that
is just about I would say main steam line strength or so and
the 1X is essentially a pristine tube, so you can see the
pristine tube is going to fail last and this is a
sequence -- I'm sorry I don't have the other slides to show
you the temperature and pressure differences but what is
happening in this case is this is one of the intermediate
pressure cases and you have some repressurizations, the
depressurizations, and there is a question about what
happens when you have pressure pulses also.
What happens on the first pressure pulse is that
you force the hot gas up into the tubes and then because
there is not much on the outside of the tubes in a
depressurized generator they don't cool off very quickly and
then what happens in the next pressure pulse is that they
are already hot and you start accumulating creep damage, so
you start seeing this stepwise behavior.
It gets quite complicated, especially when you
look at this variety of different strength tubes because of
different flaw sizes.
Now if we start talking about the 9505 case,
typically in the flaw distributions we see, whether they are
9505 flaws or they are free span flaws, you see a few that
contribute the bulk of the leakage, whether they are the
measured flaws in the generator that it might pop at main
steam line break or they are the projections through MONTE
CARLO.
It is not typically a large number of flaws that
would contribute just a little bit of leakage in the free
span that gives you the big total. It is the handful of
flaws that are contributing most of it.
If you start doing that realistically where they
are confined in the tube support plates and maybe squeezed
shut and you are heating everything up it is not clear to me
that those cracks will even open under those conditions, but
if they do we don't expect -- the main point here is we are
not expecting a 132 GPM leak value to occur.
We are thinking it is going to be closer to the
one that we know we are permitting. Originally when we were
doing these we were talking about not one but six and we
were pretty confident that something that would leak six in
the free span that was encased in tightly-clenched crud
would probably not leak one.
As the number went from six to 20 to 50 to 132, we
thought we needed to start asking the question again about
how much it leaked through the crud.
CHAIRMAN POWERS: When we talk about the cracks
contained within the top and bottom planes of a tube support
plate, I think yesterday when we discussed those cracks we
said that indeed there were opportunities for those cracks
to extend above those two planes?
MR. LONG: We don't allow that. The question is
can we always detect it, can they grow during the cycle, the
intent is to not have them do that, and, somebody correct me
if I am wrong here, but I think it's sort of immediately
reportable if it's detected to have occurred.
MR. STROSNIDER: This is Jack Strosnider.
I think what Ken Karwoski was referring to was
there's been some metallurgical studies of pulled tubes
which showed that the cracks extended slightly above the
tube sheet and I think he was pointing out that there may
have been some crud sitting on top of some of those tube
sheets, providing that environment.
There is a requirement for licensees that adopt
95-5 to inform the Staff if they detect flaws extending
outside the support plate.
Now clearly their ability to do that is driven by
the certainty or the confidence you have in the inspection
but typically the length sizing is somewhat better and also
it's my understanding when you look at the eddy current
trace you can see the edges of the support plate so you have
got some reference point there to work with so -- but at
least in terms of any significant crack extension beyond the
edges of the support plate I think we have got controls in
place so that we don't have worry should it happen.
MR. LONG: I think the next step is human error
probabilities and it's Gareth Parry.
CHAIRMAN POWERS: Am I correct, Gareth, that you
have flow in special for this extraordinary opportunity?
MR. HOLAHAN: Let me confess to having dragged him
in.
[Laughter.]
CHAIRMAN POWERS: I happen to know that he looks
forward to every one of these opportunities. He probably
will send you a note of thanks.
MR. HOLAHAN: Having dragged him in, let me soften
up some of the blows to the point of Gareth didn't do many
of the analyses that he is going to talk about and I think
he might not have done any of the analyses that we have
talked about in the last two days.
The people who did those analyses are either not
available or they don't work at those places that they
worked when they did the analysis for the Staff. Some of
the things that he is going to present to you were sort of
patched together from information that are in a number of
reports, so if the questioning gets too hard I will try to
protect him a little bit.
CHAIRMAN POWERS: Well, understand that one of the
things that we very much want to be able to respond to is
the contention that the human error probability is taken to
be 10 to the minus 3rd and we need to understand how that
number came about.
MR. HOLAHAN: I understand and I have to confess
that as an amateur PRA practitioner I did some of the human
reliability analysis on one of the earliest reports.
CHAIRMAN POWERS: Let's see. If we go through the
SME qualifications --
MR. HOLAHAN: Not even close.
[Laughter.]
MR. PARRY: With that I will basically just even
strengthen what Gary said and say that what I am really
going to talk about is very general stuff since in fact I
think the questions you had -- that accompanied the agenda
were fairly general, and if it is not what you want to hear,
please stop me and I will be happy not to tell you.
CHAIRMAN POWERS: One of the things that I very
much want to understand is this 10 to the minus 3rd human
error probability that was quoted by the DPO author.
MR. PARRY: That is not something that I can
comment on -- but what I will do I think is just tell you
the process that as an HRA practitioner you would go through
and then maybe somebody could help you to see whether in
fact in the analyses that such a process was in fact gone
through.
CHAIRMAN POWERS: Can you give me some context to
put to the 10 to the minus 3rd, what kinds of human
activities have probabilities for human error of 10 to the
minus 3rd? Nothing I do, I know that --
[Laughter.]
CHAIRMAN POWERS: I hope. Point one is on the
best day I've ever had --
MR. PARRY: So you crash your car every one in 10
times you are supposed to brake? I don't think so.
CHAIRMAN POWERS: Good recovery.
DR. BONACA: The 10 to the minus 3 was associated
with the failure of the operator to depressurize and cool
down, that step.
MR. PARRY: For what?
DR. BONACA: For a steam generator -- essentially
for a rapid cooldown caused by a steam line break on the
secondary side followed by difference size ruptures, okay,
in the steam generator tubes ranging between 100 to 1000 GPM
so a fraction of the tube to about two tubes.
I guess, just to give some background on that, it
seems as if looking at the scenario you have some indication
at some point in time that you have both a blowdown and
depressurization event and also some leakage to the
secondary side.
The time involved here in this scenario is hours
really.
DR. SIEBER: Maybe you should use two hours. The
whole event is --
DR. BONACA: No, no, that would be four, bigger
breaks. You know, this is only up to about 1,000 GPM.
DR. SIEBER: And the whole thing would be
accompanied by a lot of noise and shrapnel preventing verbal
communication.
DR. BONACA: But you have the destruction
procedures.
DR. SIEBER: Right.
DR. BONACA: The ERGs, which also include these
kind of scenarios.
DR. SIEBER: Right. And I guess that -- on the
basis of those sort of conditions, if you can convince
yourself that the scenarios, in fact, -- if the procedures,
in fact, do help you through those scenarios to the correct
actions, and the cues are fairly obvious and not confusing,
then if you have that much time to react, and, presumably,
it doesn't take that long from the depressurization, I
wouldn't have thought that 10 to the minus 3 was an
unreasonable number.
You do find in PRAs human error probabilities even
as low as 10 to the minus 5 for very protected time scales
and for things that are obvious like initiation of
suppression pool cooling in a BWR. I think where you tend
to have high error probabilities is where the conditions are
such that the cues are not obvious, or the procedures are
not helpful, or there just isn't much time.
So I would have thought that 10 to the minus 3 was
not necessarily a bad number.
MR. HOLAHAN: Can I go back historically? Not
that I want you to take away the mid-1980s calculations as
our best current thinking, but I think they do address one
important aspect, and that is that quoting 10 to the minus 3
is misleading. The analysis done in the 1980s, and the
stuff done by INEL and by the staff in the 1990s, and you
heard about some of the thermal-hydraulic analysis earlier,
those analyses are very similar from the point of view of
the thermal-hydraulic and the systems analysis, and the
amount of time available and what needed to be done.
In fact, in the 1980s, a value of 10 to the minus
3 was also used, but it was used for what I would describe
as the simplest cases, and those were the cases of a single
tube rupture with either a main steamline break or some
other secondary side failure in which the times available
for operator action were in the range of 15 to 20 hours,
okay. And those are the cases that were ascribed to 10 to
the minus 3.
And looking at the multiple tube failures, in the
range of two to 10 tube failures, times tended to be on the
order of about five hours, and those were given a 10 to the
minus 2 on reliability. And cases of 10 and more tube
ruptures, in which case the operator had actions to take
more or less on the scale of one hour, were given .5 failure
probability. So when you hear the number quoted, it is not
for the most extreme multiple tube rupture with a big
steamline break, but it is complicated.
DR. POWERS: Let me ask a couple of questions. We
have got our expert here. Maybe we deviate a little bit
from your planned presentation.
MR. HOLAHAN: That's fine.
DR. POWERS: I am looking for insight on these
numbers. One of the -- and maybe, Mario, you are the right
one to describe this a little better. At least when we look
at it, it seems to us that there are protracted times for
all small numbers of tubes up to maybe not 15, but certainly
10, that are hour times of timeframes.
We have more troubles with loud noises and
shrapnel and all kinds of things going on. But in thinking
about it, we said, gee, the cues available to the operator
to understand what is going on are perhaps least at one
tube, and if he has a long time to respond, he can easily be
confused, but they become much more clear as we move up to a
few tubes. And then, as you move beyond that, you start
losing time. So that there might be an optimum in here of
tubes.
MR. HOLAHAN: I am convinced the optimum is zero.
DR. POWERS: You are looking at a grander, on a
larger scale optimization than I am. Now, is this
completely ridiculous thinking?
DR. BONACA: No, no. In fact, I think that the --
well, first of all, yeah, what Gary said is correct. I mean
this is in reference to NUREG-1477 where we pointed out it
is between a fraction of a small pinhole probably and range
all the way to maybe a tube, tube and a half, something like
that at. And if you look at the INEL analysis, they have
made different assumptions, because they have more tubes and
they go in 10 to the minus 2 and then .5. And so there is a
consideration of time.
Second, the INEL report makes the consideration
that when you go to beyond three to four tube ruptures, the
hole is large enough that you cannot repressurize.
Essentially, the pressure comes down on the primary side and
rather than coming back to the shutoff head of the high
pressure injection, tends to stay low, and there is clear
indication that there is a hole in the system. And so the
system itself drives itself to the conditions of
depressurizing and pulling down, I mean just simply it is
going there.
DR. POWERS: It is going itself.
DR. BONACA: And now again, even for those
scenarios, you have hours of time still to take some action
and, clearly, if you don't take action in two, three hours,
then you are going to go toward depletion of RWST. But the
procedures, if you look at the ERGs and you read them over,
there is a lot of consideration of that concern of RWST --
RWST depletion.
So they are not moot about that, they are talking
about the need of maintaining subcooling, but also to
prevent RWST depletion. And so you don't pump water for
hours and the operator simply is unaware that he is
depleting the RWST. In fact, he is going to be very
concerned about that.
And the other thing is that, which is encouraging
to me, is that the ERGs speak about the possibility of going
to RHR in a saturated mode, which means they are informing
the operator even during the training that he may not be
able to recover subcooling. But he then can -- which
implies that he has a large hole in the system. Okay.
So there are, you know, there is a lot of
information in the ERGs to be encouraging.
Now, the only thing that is confusing, and I want
to point out is that, if I remember, when you have a
steamline break, you have containment desolation, and you
have also -- I believe you have loss of the air ejector.
MR. HOLAHAN: Yes, that's correct. Yes.
DR. BONACA: Okay. So there is lack of some
indication there to make -- so that may delay at the
beginning his determination that he has a hole in the
system. But I don't think these numbers, I mean are that --
are reasonable. 10 to the minus 3, again, it is reasonable
in a scenario that lasts for 10 to 20 hours.
DR. POWERS: Well, my recollection is that we saw
a discussion. We had -- I mean it was a discussion I think
of perhaps the Halden reactor, where you had poor
performance despite these times and whatnot. I mean do we
understand why that is?
DR. BONACA: Well, first of all, I think -- I am
not sure the presentation really represented the situation
today where the ERGs are an established symptom oriented set
of procedures. I daresay that in the '80s, I would not have
the same level of confidence at all, because there was no
structured process to recognize, for example, this potential
for rapid cooldown and steam generator tube rupture. But
the ERGs recognize that very explicitly because they are
telling you how to get there.
And I don't know about the Halden project, if it
is recent, and I am not sure that the operators represented
there had, in fact, the helpful procedure structure the way
the ERGs are.
DR. SIEBER: I think there is another factor, too,
because you would end up in some kind of a callout status at
the plant, and you would have more help than you could shake
a stick at, including the technical --
DR. POWERS: That was universally recognized as a
bad thing.
[Laughter.]
DR. SIEBER: Well, nowadays it is supposed to be
organized and structured. And what you don't want is a lot
of people running in and out of the control room. On the
other hand, you have the ability to have turnovers. You
have the ability to do calculations. You have the ability
of innumerable people to critique and watch what is going on
and provide technical assistance.
The other thing that is not on that sequence is
there is a lot of other things that happen, because if your
power which causes the accident conditions, you get a
turbine trip or reactor trip, you have about 35 things that
you have to do to respond to that, and they are going to
open up safety valves, which make almost as much noise as a
break someplace in the steam system.
If it is inside the building, all your fire alarms
are going to go off, okay, like happened at Surry. And so
you are going to have enunciator lights and computers
reeling out tons of stuff on CRTs. And if that is
accompanied by a tube rupture, and you don't have in control
room N-16 monitor outputs, you are going to have a problem
recognizing that right away, because the reaction of the
parameters on the reactor coolant system which the operator
begins to monitor is the same for steamline break as it is
for a tube rupture for that first increment, until all of a
sudden, because you are going to go pretty far down on
pressurizer level and pressure is going to come down. The
plant is going to cool off pretty severely.
So it isn't until you are into that a little bit,
and you get that blowdown and the cooldown, you can tell
that, uh-oh, I am on a different path than what I would
expect.
DR. BONACA: That's right.
DR. SIEBER: With N-16 monitors, which are
required and aren't Reg. Guide 1.97, you can pick it up
pretty quick.
DR. BONACA: The last comment I would like to make
about that is that, you know, 10 to the minus 3 is always a
very hard number to -- you know, it is a very small number.
But the other comfort I got in reviewing this material is
that it comes out to an increasing CDF of 2 in 10 to the
minus 6, and I thought, what if it were 1 in 100, it will
come 2 in 10 to the minus 5.
So that gave me some comfort than even with
significant uncertainty applied to it, I would still get a
relatively small increase in CDF.
MR. BALLINGER: I need to get something squared
away in my mind. In the case of IP-2, the staff assigned a
probability of failure of .1 for that event, and I see 10 to
the minus 3 here. Operator failure.
MR. HOLAHAN: Failure to do what?
MR. BALLINGER: Failure -- now, that is what I
want to get square away? I mean Jack's -- well, it was
here.
MR. HIGGINS: I think it is important to realize
that we are talking about many different sequences here,
Ron, with all the different things, because over the last
two days we have talked about -- I mean we have gone through
the spontaneous steam generator tube rupture. We have gone
through the various accident induced ones that delta P
inducted. And all of those have somewhat different operator
actions associated with them considering the timing and
considering the actions and the stresses that John was just
describing, and those are all going to have different HEPs
when you do the calculations. So it is very much too
simplified to just say that 10 to the minus 3 is the number
used in these analyses.
DR. SIEBER: Another factor is that Reg. Guide
1477, I guess it is.
DR. POWERS: NUREG.
DR. SIEBER: NUREG. Really looks at the accident
as -- since the Reg. Guide 95-05, assumes that the tubes
don't rupture and just leak. It follows the simple event
tree of a steamline break, which is much simpler than having
these two events going on at the same time. And so the
analysis in 1477 may be justified because the accident, the
event tree that you are analyzing is simpler than one that
has these two accidents going on.
Actually, the question is, does the steam
generator hold up? And if it doesn't, it leads you into
another sequence which hasn't been analyzed here.
MR. LONG: What is 1477 was intended to look at
the DPO issue of a large amount of leakage due to cracking
that was in the freespan. So if you look at the event tree,
there is no conditional probability that leakage will occur.
That was just put in as one.
DR. BONACA: As one, yeah.
MR. LONG: And so it didn't really appear in the
tree. And then the intent was to try to deal with the
combined event. And, initially, we simply lifted the human
error probabilities from NUREG-0844 that Gary was talking
about earlier, and we went through some analyses to try to
figure out where we would leave them to be, with some
additional effort that I described yesterday, to some
degree, at least up to the point of what the inputs where.
And eventually, I believe, in 1570 we used 10 to
the minus 2 instead of 10 to the minus 3. So we did shift,
but we were still dealing with moderate primary to secondary
flows, not, you know, tens of thousands of GPM, but maybe a
thousand GPM or multiple hundreds of GPM for those events.
MR. PARRY: I think, though, the key really is for
them to be able to understand the status of the plant as it
-- particularly with the failure of both the secondary and
the primary side and whether the procedures will lead them
down that path.
I think initially, the -- I only know the
Westinghouse system, and that's from a few years back. I
guess initially there would be an E2, which would be the
steam line break from the generator and then maybe
transmission into E3 or even E1. And eventually, they would
end up probably doing the right things.
MR. BONACA: Yes. I mean, I didn't see anything
--
MR. PARRY: They all lead down the same path.
MR. BONACA: Yes. It will lead down the same
path. I believe tougher is going to be a small leak because
you have a steam line break, you don't know that you have a
small leak. But you have plenty of time to --
MR. PARRY: Right. To compensate for that.
Actually, in a sense, you cut straight to my last viewgraph
with your talk, so I'm really not sure it's worth going
through what I've written here because I think we have
covered the issues that -- yes, there's a possibility that
there is a confusion factor, and that's something that has
to be taken into account. The more confusing it is, the
less likely the likelihood they will succeed.
MR. HOLAHAN: I think there has been some
misunderstanding in the past on this point, a
misunderstanding that the staff had intended to use the
human error probability of ten to the minus three for some
extreme multiple tube rupture cases, and that has never been
done. So I think it seems to me that the real issue is not
how the operators would respond. No one is going to give
them credit for handling 100 tube ruptures with a main steam
line break. The real question is how likely is such a thing
to happen? Are there real mechanisms that would allow such
a thing to be sufficiently likely that they need to be
modelled.
MR. BONACA: The other thing that I think is
confusing somewhat is that the objective has always been one
of, you know, not emptying the ARWST. But as Dr. Ward
pointed out this morning, then there are hours before you go
to core uncovery, about four hours, and so it seems to be
very unlikely to think of an event of this kind evolving to
the point where you're emptying the ARWST and then you sit
there for four hours without doing anything. I mean, I
think in this comprehensive scenario, there are many
opportunities to take action and --
MR. PARRY: Yes. I mean, isn't there the
contingency to refill the RWST called out in the procedures
as well if you don't have anything in the sumps.
MR. BONACA: That's right.
MR. PARRY: Those are things you can do, and
that's probably not taken into account in these analyses is
my guess.
MR. BONACA: That's right, as well as, for
example, connections already existing with other tanks on
the site.
MR. PARRY: Right.
MR. BONACA: Many sites have additional RWSTs
available for make-up.
MR. HOLAHAN: Since I've already confessed to be
being a amateur HRA analyst, I would like to add three
thoughts.
The issue about operating experience showing that
operators didn't handle the events very well I think all
relates to the design basis issue of quickly isolating the
generators in the time frame of 30 minutes, and I think
those are valid criticisms, that the traditional use of 30
minutes is, in fact, not so easy for operators to figure out
which generator has the leak and basically to isolate that
generator in 30 minutes, because, in fact, operators,
although they figure these things out, the real process of
acting is more deliberate than the analysts assumed 30 years
ago.
Second insight is, at least from the NRC's end of
the phone calls, I've seen a number of events and many, many
drills, and there's a great deal of sensitivity to radiation
anywhere outside the reactor coolant system, and I think one
of the things we're talking about is, you know, an operator
having knowledge that there is a steam line break and a tube
rupture and radiation signals from around the plant I think
would, especially over the time frame of hours, would be
something the utility would be very sensitive.
Thirdly, the NRC operations center two floors up,
if we're talking about 15- and 20-hour scenarios and going
to core melt, I would have to think that we would have
failed on our end in figuring out what in the world was
going on in those plants, and as the director of the reactor
safety team in the operations center, I have a hard time
saying that we wouldn't figure it out on our end.
MR. BONACA: I would like just to add that I agree
the 30 minutes objective right now is one that seems to me
that is somewhat -- the operators almost because it's a
requirement that has to be met. But if there are some
complications there, they may not pay attention to those
because they're so focused on equalizing pressures between
primary and isolated steam generators within 30 minutes,
which is very challenging for them to do.
CHAIRMAN POWERS: Is it your perception that this
evaluation that was done for the Halton staff -- when it
says poor, is poor relative to a 30-minute time window which
seems to be a completely arbitrary sort of thing?
MR. HOLAHAN: It's not arbitrary; it's part of the
design basis dose calculation that Jack Hays showed you
yesterday as leading a small fraction of the part 100. But
from a severe accident point of view, it's irrelevant.
CHAIRMAN POWERS: Okay. Well, I guess I'm looking
at a design basis accident point of view right now.
MR. HOLAHAN: What I would say is from a design
basis point of view, the steam generator tube rupture and
dose calculations have many, many conservatisms. We once
calculated about four orders of magnitude of conservatisms
in the dose calculations, okay? And I think we talked about
iodine spiking and looking for the 95th percentile and the
meteorology 95th percentile.
Well, the one thing in that sequence that's not
very conservative is the time to isolate the generator,
because I think 30 minutes is certainly possible, you know,
but experience shows that 45 minutes or an hour is more
likely to see what happens.
But I think if you see that in the context of the
overall conservatism of the design basis calculations, it
doesn't bother me very much.
CHAIRMAN POWERS: Design basis are always very
confusing to me. I mean, there seem to be times when we're
lenient and times when we're not, times when we invoke risk
and times when we don't. Clearly in the design basis
analysis, by the time the day is over, we have no idea what
the total level of conservatism that you compose because it
shows up in multiple places. But you also have the same
problem when you start granting leniency, that it doesn't
bother you very much on these things. You don't know how
much of the margin you have taken away.
MR. HOLAHAN: But in this case, it doesn't even
bother me very much with respect to meeting the part 100
dose guidelines.
I understand it's a little different when you say
we're going to shave design basis margin because I don't
think the risk implications are very important. In this
case, I think the exact time of steam generator isolation
isn't really all that critical to meeting part 100
guidelines.
CHAIRMAN POWERS: I think I understand.
Do you have other points that you --
MR. PARRY: Not really.
[Laughter.]
MR. HOLAHAN: I would just like to summarize on
one point. The numbers I read you and that Steve said had
been picked up are at least 15 years old, that when we redid
some of the analysis in the 1990s, we rightly thought that
they should be re-looked at, and INEL did some human
cognitive reliability analysis and came up with some
approach.
But even when they did those analyses, they
identified them as screening type analysis and they thought
that some additional work ought to be done to, you know,
refine the answers.
So I think we're not saying that we know or have
really solid information on human reliability. I agree
completely with Dr. Bonaca's observation that you can do
some sensitivity studies and change the answers and see that
it's not all that critical if the values aren't quite ten to
the minus three, and they're certainly not ten to the minus
three, nor have they been claimed to be ten to the minus
three for the most severe cases that we've talked about.
CHAIRMAN POWERS: Gareth, I think I want to ask
you a question. It's going to be very difficult for me to
put forward. It's not a question you're going to want to
answer.
MR. PARRY: Then I won't.
CHAIRMAN POWERS: I'm going to plead passionately.
We have this design basis time window of 30
minutes in which we would like the operator to identify the
leaking steam generator and isolate it. We are told by Mr.
Holohan that this is a challenge for them, that in fact a
better time period for doing that isolation process might be
45 minutes to an hour. Jack has described to you a chaotic
situation in which there are lots of alarms going off and
whatnot. At the same time, we do have a pretty good set of
procedures.
From your vast storehouse of experience and
knowledge on these subjects, what would you guess the
probability that the -- I don't want to call it an error
probability -- the probability that an operator would fail
to complete this task within the 30-minute time frame?
MR. PARRY: You're right, I wouldn't want to
answer that question.
CHAIRMAN POWERS: But I'm going to plead so
pathetically.
MR. PARRY: And the reason I wouldn't, I think, is
because it's so dependent on the details of the procedure
and the training.
But let me give you one little insight, that if we
were analyzing -- typically if you're analyzing spontaneous
tube ruptures and you are concerned about the isolation of
the generator, the success criteria in most PRAs as I
understand it, or certainly the ones we used to use, were
not 30 minutes, it was before the steam generator
over-filled, which typically would be on the order of an
hour depending on the size of the leak.
So -- and for those -- for that particular step in
the procedure, and just the isolation, I think -- I'm trying
to think back. Typically we would probably have used an
error probability of the order of ten to the minus two. But
at that point, then it becomes a contained accident. And
the worst case is if they don't do in that time, then we
have to go down to RHR.
So those scenarios, I think the error probability
for that simple single tube rupture type scenario I'm pretty
sure was a lot less than ten to the minus three because of
the length of time available.
MR. HOLAHAN: Could I add something? I just
wanted to add something to that.
Before I came, I went through -- we were doing
some work for the STP process for the NRC as far as
developing the risk-informed inspection notebooks and
developing the operator actions and the credit for those in
those, and I went through and looked at some of the steam
generator tube rupture related human actions from IPEs and
for the PWRs, there were two that were important. One was
this early isolation of the ruptured steam generator and the
other one was depressurizing the primary, and they both
typically run around ten to the minus two. I've got some
data here from -- I don't know -- maybe 30 plants, not all
of which have clearly identified HEPs that you can extract,
as Gareth knows. But I would say in general, they average
about between 1.0 and 2.0 times three to the minus two for
each of those actions separately.
MR. PARRY: Now, what you said about
depressurizing the primary, you're talking about
depressurizing to RHR entry conditions. Is that --
MR. HOLAHAN: Right. Depressurizing it below that
of the secondary, not all the way to RHR.
MR. PARRY: Okay. Okay. Just to stop the leak.
MR. HOLAHAN: Right.
MR. PARRY: Okay.
MR. HOLAHAN: Right.
MR. PARRY: Okay.
CHAIRMAN POWERS: One of our speakers earlier in
the week presented a -- I guess his assessment of the
performance of various operational teams during the course
of a spontaneous steam generator rupture event, and I was
trying to find it, but my recollection is that it's a litany
of delay doing this task, doing the other task, delay doing
the third. Is that coached, these numbers, that you get ten
to the minus two human error probability? I mean, it's
funny because, I mean, it's a time window. The guy can do
it successfully in 35 minutes. Do I count that as a failure
because it wasn't 30? I mean, it doesn't seem right to do
that.
MR. HOLAHAN: Not in PRA space, you wouldn't.
CHAIRMAN POWERS: Not in PRA space, but we're in
design basis space.
MR. PARRY: No, we're in PRA space.
MR. HOLAHAN: Yes.
MR. BONACA: I can't remember exactly. You
remember the time frame for those tests?
CHAIRMAN POWERS: They weren't tests.
MR. PARRY: Yes, they were --
CHAIRMAN POWERS: These were events and they
extend from the early '70s up until just a few months ago.
They span quite a range.
MR. BONACA: Yes. First of all, I would separate
time. I think that after --
CHAIRMAN POWERS: Well, the story was consistently
the same. It was always delay doing something, and my
recollection of IP2 was there was a pretty good story there,
too.
MR. PARRY: Did any of those events lead to
over-filling the generator?
CHAIRMAN POWERS: I believe there was one of them
at least that did lead to over-fill of the generator.
MR. PARRY: That one I would count as a failure
for the isolation.
MR. HOLAHAN: I would consider it a design basis
failure. Among other things you would have released water
as opposed to steam and so partitioning and a lot of other
things in the analysis don't come out right.
CHAIRMAN POWERS: You would jump all over Ginna?
MR. HOLAHAN: I believe I did.
[Laughter.]
MR. LONG: I think part of the point here is that
Jim has talked about the human errors for our failing to
isolate when the -- well, in your case I guess it was the
overfilling, it was the IPEs.
That is not the whole step to core damage though.
If you look at the way the rest of the logic goes, there's
typically another human action in there or some other
equipment failures that you have to use to get the core
damage so if you take the product of the human errors, that
gets you all the way to core damage in that particular cut
set that is just pretty much all of the operator's fault.
The number typically comes out more like 10 to the
minus 4 as the total product, maybe lower depending on the
PRA.
MR. PARRY: That's right.
MR. LONG: That was the kind of number we were
trying to capture when we did the event trees for 1477 and
the sort of event lists for 0844. It was the total process
so that top event was operator fails to pressurize, cool
down RHR.
In that regard I think we are being somewhat more
conservative than what you would see for the spontaneous
rupture for the overall human error.
DR. SIEBER: I guess what I am struggling with, I
have got 10 to the minus 3. I have a historical inventory
of events that I will admit goes from the Dark Ages to the
Modern Day.
MR. HOLAHAN: Zero for 10 core melts.
CHAIRMAN POWERS: But one that excited the
esteemed Holahan and got him agitated and he considered a
failure.
MR. PARRY: No, a failure to isolate the
generator.
CHAIRMAN POWERS: Now he considered it a failure.
MR. LONG: It was a failure to prevent overfill, I
think was the issue.
CHAIRMAN POWERS: I think it was a failure to
prevent release of radioactivity to the outside.
MR. HOLAHAN: That is before I became an amateur
PRA expert.
[Laughter.]
CHAIRMAN POWERS: It is not entirely clear that
that is a step forward on the evolutionary path.
DR. BONACA: Let me just say a couple of things I
would like to say about that.
First of all, again as I said yesterday, steam
generator tube rupture within the constraints of what they
are supposed to with the objectives, the 30 minutes, some of
the most challenging sequences, because the time is short.
Certainly they are not going to have any help very much.
That is control room delivery issues. Many of
them are dealing with other issues. For example, because of
spurious safety injection actuation many of them are running
with their block valves closed on PORVs on some of them, and
then that complicates the ability of depressurizing and all
this kind of stuff.
So the 30 minutes becomes a real difficult time,
okay? Now here when I was looking at these other scenarios,
which is very different -- you have a depressurization on
the primary side and you have a tube rupture in addition to
that, both of them are helping in the direction of going
towards a target which is the one of depressurizing and
cooling down within hours.
That is a different story because when it passes
30 minutes you are going to have, with an event like this
you are going to have all kinds of help coming down to the
control room. Now hopefully it is not all confusing, the
help, but people are going to begin to see things and there
are signals around the site telling things so even if there
was a guy who absolutely misunderstands the event, the
others will not.
I mean that is -- that was one consideration I had
in the sense of how confusing is it going to be, how
ambiguous is it going to be.
The other issue is -- again, I don't want to
minimize that -- I said to myself what if it was 1 in 10 to
the minus 2 and that still would get to a number of two 10
to the minus 5 for CDF so the contribution was still
acceptable with the significant error range that because
again I mean there is an uncertainty there --
CHAIRMAN POWERS: The truth of the matter is that
two times 10 to the minus 5th is not what I would call
inconsequential.
I mean that gets me interested at least in the
sequence.
MR. LONG: That was with guaranteed massive
leakage.
CHAIRMAN POWERS: I keep coming back -- I keep
coming back -- I know the consternation of most to the
design basis issue because I think that is the issue I have
to confront.
Suppose that I said that I will forgive the 30
minute window. I don't care. That's somebody else. What I
really, really want to do is I want to prevent the release
of enough radioactivity to get the younger Holahan prior to
his exposure to PRA excited about the radioactivity release.
Why would I not be justified in saying that the
failure to keep Holahan happy criterion is a .1 probability?
MR. LONG: Based on the empirical evidence?
CHAIRMAN POWERS: Yes, the empirical evidence yes.
DR. BONACA: For steam generator tube rupture?
CHAIRMAN POWERS: Yes, spontaneous steam generator
tube rupture. I have one that I know for sure got him
upset.
MR. HOLAHAN: Which didn't exceed Part 100.
CHAIRMAN POWERS: I understand. I understand --
still got you upset. I mean it got you interested.
MR. HOLAHAN: It's cold in --
CHAIRMAN POWERS: It's not terribly cold, it
snows.
DR. BONACA: I wouldn't disagree with that
estimate for the steam generator tube rupture.
MR. LONG: I think people have modified procedures
a lot since then so you might get back into Gareth's
expertise by trying to figure out what the new probability
is if you think it has changed. That is the issue.
CHAIRMAN POWERS: I think that's a very fair
statement.
MR. HIGGINS: Let's say it is and it may very well
be .1 to do the thing you just described.
Is that an issue? I don't think so.
MR. HOLAHAN: Not necessarily. As a matter of
fact my sympathies at the moment are with Dr. Bonaca.
I mean if we had to do it over again I would say
we have got to be more realistic in the overall calculation
of, you know, dose and consequences of steam generator tube
rupture and put less demand on the operator and a little
more demand on the meteorology and we may very well meet the
same goals in a more appropriate fashion.
MR. LONG: One thing -- I'm sorry, Gareth, go
ahead.
MR. PARRY: No, I wasn't going to say anything.
MR. LONG: One thing that we should mention before
we get off the subject is that the Office of Research has a
program and I think you have heard of that called Athena, to
look at errors of commission and omission and procedures,
and I think one of the things they are doing this week that
is making it hard to get the right people in the right place
at this time is to start looking at steam generator tube
rupture issues with that new process.
Also, I want to point out that Indian Point --
Consolidated Edison has proposed breaking the steam
generator tube ruptures, at least the spontaneous ruptures,
into two categories, sort of like small and large LOCAs or
small and large tube ruptures, the splits being kind of
plant-specific, the bottom of the small being that your
first charging pump has run out of capacity and you have to
do something to add charging and the top of the small being
you have no more charge to add -- you have to go to safety
injection and then the safety injection is the -- onward is
the larger sizes.
I think there's some benefit to that because I
think the human errors are probably different and the
opportunities for making it worse are still there while you
are in the lower leak rate --
CHAIRMAN POWERS: I think the opportunities to
make it worse is the advantage of doing that.
I mean I see it as an advantage.
MR. LONG: One of the other things -- we haven't
mentioned it yet -- but if you have the secondary side
failure first, one of the things the operators are worried
about is the cooldown rate, and there's a lot of competing
things in there -- keeping the core covered, keeping
subcooling margin, trying not to get your cooldown rate to
be too large, and they'll sometimes try to heat back up real
quick because it is over an hour the way they see it and
they are trying to put this all together with the Athena
program, so I think there is some opportunity for re-looking
at this and maybe coming out with more of a consensus on how
to do these things, because as I pointed out earlier, just
in the IPEs there were almost four orders of magnitude
difference in the result of the way people were applying the
logic to just the spontaneous rupture in the industry right
now.
That is not a very good, firm basis. My next
slide is on uncertainties and that certainly is one of them.
MR. HOLAHAN: I wanted to tie this issue back to
the design basis. I think you said you want to wrap that up
somehow.
It seems to me that there are a number of issues
here which could use a re-look by the Staff at how the
design basis steam generator tube rupture is treated.
Frankly, I wasn't completely happy with our
discussion of iodine spiking in that I think the story
wasn't entirely convincing, although I think the licensing
basis that we have used is reasonable but I don't think we
told the story in a convincing way.
I think there are a number of conservatisms in the
design basis steam generator tube rupture that are not
necessarily serving the public or licensees very well, which
in my mind makes it a good candidate for risk informed
re-look.
In that context I would say you could look at
realistic iodine spiking. You could look at the demands you
are putting on the operators and what makes sense and what
is counter-productive.
You could re-look at the conservatisms in
meteorology and other issues and I think today we could come
up with more sensible design basis requirements for steam
generator tube ruptures than we have inherited over the last
30 years.
CHAIRMAN POWERS: I am pretty sure I agree with
you.
[Laughter.]
MR. HOLAHAN: And if a committee were to recommend
that to me, all I would have to do is prioritize it with our
other risk-informed activities.
CHAIRMAN POWERS: I understand that the committee,
this committee, gathers facts and provides information to
the ACRS. The ACRS will in turn make a recommendation to
the EDO.
MR. HOLAHAN: I certainly wouldn't want to
influence that process.
[Laughter.]
CHAIRMAN POWERS: And let me assure you you
haven't.
MR. HOLAHAN: Thank you.
[Laughter.]
CHAIRMAN POWERS: I think I understand better my
design basis human nonconformance probability. I think we
have lots of fertile thinking on the severe accident side of
this.
I, myself, find very attractive this idea that
there are gradations in that error probability that are not
linear and with the magnitude of the break kind of
fascinating.
Is there anything else you need to tell us?
MR. PARRY: No, but if you are more interested in
human reliability there is a graduate course they give at
the University of Maryland --
DR. KRESS: Who is teaching that?
CHAIRMAN POWERS: Anyone I know?
MR. PARRY: Possibly.
CHAIRMAN POWERS: Could I get a good grade?
[Laughter.]
MR. PARRY: That depends.
MR. HOLAHAN: I can tell you, I am not willing to
take the test.
MR. LONG: I guess the next subject on there was
uncertainties in the risk assessments.
Shall we plunge ahead?
CHAIRMAN POWERS: Sure, please.
MR. LONG: I think we have talked about these to a
large degree. There are sort of three areas that I want to
talk about. The human error probabilities. I think I won't
spend any more time talking about the uncertainties in
those.
We have talked about the NDE detection of flaws,
and I think you have seen the POD in that. I will mention a
couple of things that came out of some reviews of license
applications.
Then there are the tube strength estimates based
on the NDE characterizations of the flaws.
When I wrote that slide, I think I left off
thermal-hydraulic modeling uncertainties.
MR. POWERS: I didn't think there were any. I
thought thermal-hydraulics was a well established field of
an exact science.
MR. CATTON: It's considered to be a mature
science, but not exact.
MR. POWERS: It's only the participants that are
mature.
[Laughter.]
MR. KRESS: Geriatric science.
MR. LONG: When we did NUREG-1570, we tried to do
sensitivity studies on the various things that went into the
risk assessment. What we really figured out was it looked
like we were very sensitive to what the flaws were. If you
took a different flaw distribution, you got a different
answer. We were very sensitive to what the temperatures
were on the tubes at least relative to the surge line in
terms of heatup rate of the tube in competition.
We were sensitive to whether you had cutting from
small flaws or not. In other words, the cutoff size of 0.25
and how much you worried about the small flaws. We knew
that, and we reported that in the report.
Then, as we went forward and tried to apply this
later on, we found some other things. In particular, with
the thermal-hydraulics the RELAP/SCDAP output is the
temperature of one assumed to be representative of the heat
transfer hot tube. I'm not sure exactly what that means in
terms of average over the tube sheet for the different tubes
that are carrying the flow, but when we do these
calculations we need to know what the hottest tube is. We
want to know if either the pristine tube or a tube has sort
of an undetectable expected amount of degradation in it
since the tubes are to some degree aged and there are some
things that are on the order of 20 percent through wall you
probably just can't find.
We don't really have that. There are varying
opinions as to how close we are to that with the RELAP
number. If I talk to some people, I hear, well, we're very
close. If I talk to others, there is some concern that we
are not very close at all.
So that is the beginning issue.
The next part of it is, if somebody tells me I
have a few flaws and they are distributed somewhere in the
hot leg side of the tube sheet, I don't know, first of all,
if they are in or not in the hot part of the bundle. If I
am told that the hot part of the bundle is 53 or 35 percent
of the bundle, at least I have a statistic I can start using
to try to get a probability that one of my bad flaws is
within that region.
But within that region there is quite a variation
in tube temperature. So I have difficulty in trying to
figure out what the probability is that my weak flaw is
going to my hottest tube or a tube that is at least hot
enough to cause it to fail before the surge line.
I tried in the Farley analysis about a year and a
quarter ago to squint real hard at the distributions of tube
temperatures like I showed you before and tried to get some
areas that I thought were hotter by a certain amount than
the temperature that RELAP predicts, and for that matter,
there has to be some that are cooler as well to make that
some sort of an average. I purposely didn't write down the
details because I didn't think I did well enough that I
wanted anybody to just copy it. In the NRC, if you are not
careful, it will just be copied by the licensees from then
on because it's something that they think we are going to
approve since we did it ourselves.
I noted that if I scaled the difference in the
tube sheet temperatures just from what I could see from
variation in the tube sheet, if I took the delta between the
cold and the hot as the scaling parameter and then looked at
the fraction of that delta, I get a different answer than if
I took the delta from the hot leg to the tube sheet and
looked at the variability.
So there is a real issue here about how do you get
a distribution of temperatures on the tube sheet.
MR. CATTON: I would agree with that.
MR. LONG: I knew you would, but I know some
others won't.
So it basically comes down from the mixing of the
countercurrent flows and what we can do with those.
MR. CATTON: This is the same exercise that I went
through a few years ago to the same conclusion.
MR. LONG: We have talked about some beginnings to
try to get more information on that. There is some question
about how far we can go without doing physical studies.
Right now NRR has asked Research and Research has responded,
and we have said, yes, that looks like a good start. The
question is, will we get to what we need to do these things
adequately for licensing purposes.
We have talked about the effects of leakage on the
tubes. We don't know where that really starts becoming
important.
We also talked about sort of the non-stylized
accidents where you have leaks of different sizes in
different places and the RCS and what effects that may have.
It really complicates the picture quite a bit.
We also have a concern that we seem to get
consistent differences between MAAP calculations and RELAP
calculations. Of course there are people who wrote one that
are throwing bricks in the direction of the guys that wrote
the other. Mark Kenton has been doing a fair amount of work
to try to figure out if he can make MAAP look the same as
RELAP. One of the things that he has picked up is he thinks
he sees an importance in radiative heat transfer between the
fluids and the walls.
MR. CATTON: What is the fluid?
MR. LONG: At the point he is doing it, it is high
pressure, high temperature steam.
There is also the differences that the licensees
are giving us calculations with one code; we are using
another code, and they don't tend to predict the same order
of stuff failing necessarily, much less the same timing or
the same temperatures.
So we feel there is a fair amount of uncertainty
here, and it makes it difficult to do an analysis and then
to take that into the decision making process.
I will go a little bit further than the slide was
intended to go and say, when I had to do this for Farley, I
sort of had an option of telling Farley that their
application really didn't address Reg Guide 1.174 or to
recognize that we really hadn't ever put out any guidance
although we had been asked for it for years and to go ahead
and follow the other guidance I have, which is to say, if
you can reasonably figure it out for yourself, go ahead and
do it. So that is what I tried to do to see how far we
could get. Because Farley was so much like the Surry plant
we have studied for a couple of decades, I thought that was
a good basis to make the attempt.
Other things that are pretty uncertain come from
the creep model for the RCS components. We are assuming
infinitely long thin wall tubes. Maybe the steam generator
tubes kind of fall into that category. But if you start
looking at things like the surge line, which we hope will
fail first, it has a lot of angles. It has restraints on
its growth. There are welds which are probably not perfect.
So the destructive effects may not be the ones we are
modeling, and if we are lucky, maybe it will fail earlier
than we model.
MR. POWERS: When you model the creep rupture in
things like the surge line, do you use damage accumulation
in the model?
MR. LONG: Yes.
MR. POWERS: Do we have damage accumulation kinds
of data for things that have to get that hot?
MR. LONG: I think somebody has failed a surge
line in a test, right?
MR. BALLINGER: This is stainless steel, right, or
is it carbon steel?
MR. LONG: It depends on which plant you are
talking about.
MR. BALLINGER: There is a lot of data for
stainless steels in these temperature ranges from the fusion
program, but not carbon steel.
MR. MAYFIELD: This is Mike Mayfield from the
staff. Surge lines are going to be either cast or wrought
stainless. Nobody runs carbon on the surge lines.
MR. BALLINGER: There probably are a fair amount
of data.
MR. MAYFIELD: We've broken them, literally a
surge line we got from a canceled plant, but it was at
normal operating pressures and temperatures rather than
these elevated pressures and temperatures. That is one of
the things we have been talking about doing in this
additional work that NRR asked us to do, to look at the
elevated temperature response.
MR. CATTON: Where does it fail?
MR. MAYFIELD: We were intentionally flawing the
pipe.
MR. CATTON: Is it the pipe itself that failed?
MR. MAYFIELD: Yes. Where it is going to fail is
where you have a crack in it, which in this case was in a
weld.
MR. CATTON: How far away from the hot leg is this
failure?
MR. MAYFIELD: This was in a straight piece of
pipe. It's wherever you put the flaw.
MR. CATTON: The surge leg comes into the hot leg
where everything is very thick and welded in there. It must
be really tough to figure out when it's going to fail.
MR. LONG: Also, if you take a look at the way we
model some of the hot legs. If we model for plants that
have stainless steel hot legs, we may model the safe end to
the vessel. The question is, is that really long thin wall
pipe at that point that is constrained at one end by a weld
of more capable material and on the other end by a very
thick vessel?
MR. CATTON: It makes this crossover even more
uncertain, doesn't it?
MR. LONG: If you start looking at the short,
complicated shapes, I don't think we are modeling those very
well at all. We are using a creep damage accumulation model
as if it's a thin wall pipe at the temperature that is the
median temperature of the full wall thickness, if I remember
correctly.
MR. CATTON: That's kind of a heat model for
RELAP, isn't it, just a chunk of metal with resistance to
the center, and the capacitance?
MR. LONG: Not knowing any better, I'll say yes.
We talking at some point about the potential for
the cracks eroding further by the flow going through them.
That doesn't look like a problem with recently acquired
knowledge. It certainly did sometime ago when we did the
last licensing applications that involved this. And the
same with cutting where we were using the 0.25 inch as
essentially a 0.25 inch through wall segment was equivalent
to primary and secondary failure.
So there are a lot of things going both ways,
conservative or non-conservative. We have had applications
claiming that the tubes wouldn't fail during severe
accidents even with the cracks and we have had applications
claiming the tubes would always fail with cracks they
couldn't detect during severe accidents. In either case,
the delta LERF is zero for what they requested. It makes it
very difficult to go through this and say we have done
everything in a conservative manner, because then somebody
can turn around and get the delta LERF by always failing,
and everything it assumes is now non-conservative.
MR. POWERS: One of the issues I think you pointed
out earlier is that when you have this steam generator with
natural circulation flow you have a temperature distribution
among the tubes of the steam generator. You look at them,
and you say, gee, I think these things can bow the tubes in
the hot zone or something. Is that what you were thinking
of?
MR. LONG: I wasn't saying they would bow towards
the hot zone. I was saying that if you have a large bundle
of tubes with a smaller batch of them at a much higher
temperature than the others, they would try to elongate,
especially if they are crimped into the support plates like
they would be with drilled holes, but there are other plants
where they are quatrefoils, or whatever. Some of them are
going to try to get longer than the ones on the periphery,
and for that matter, the shell structure. What we think
they would do if they are locked is bow. If they are not
locked, we are probably not granting credit for confinement
for degradation. In other words, if it's not a drilled hole
support plate, we wouldn't be giving them credit for the
drilled hole support plate, and we wouldn't have flaws that
would be growing to a free span. For instance, Arkansas 2
is a CE plant. It has an egg crate type of support plate,
and we treat those flaws as if they are in the free span.
There are a few things that I am going to talk
about on uncertainties further on, but I guess one thing I
should mention is for Farley I tried to integrate these
uncertainties as best I could to get one parameter for
decision making purposes. For instance, Charlie gave you
thermal hydraulic temperature uncertainty of about 50
degrees plus, and the way I would model that was to put into
the Monte Carlo process an uncertainty that was plus or
minus 50 degrees. I did it with a Gaussian distribution.
When I do things like that and I am getting beyond the data,
I will cut the Monte Carlo at the wings, so that if I am
viewing something that looks like 5 percent to 95 percent, I
will not let the Monte Carlo go out to three times that
value with some real scarce frequency.
MR. CATTON: Why do you give it a Gaussian
distribution when it's so uncertain? Shouldn't you give it
a uniform distribution? Isn't that the way the rules go
when you don't know it's equal?
MR. LONG: I didn't know those were the rules.
MR. CATTON: I don't either.
[Laughter.]
MR. CATTON: I'm just a thermal hydraulics guy.
MR. POWERS: When you put things on a Gaussian
distribution, you do need to be normalized.
MR. LONG: When we say we normalize, I am
basically putting 100 percent of the area under the
distribution. If I find something outside that, I'm just
choosing another one and going through.
MR. POWERS: If the number is outside, you just go
back and choose another one?
MR. LONG: Yes. It's not right perhaps, but given
that a flat distribution might he right too is wrong.
MR. POWERS: I would have funny results if I did
that. The check sums wouldn't work out. The probability
within that is one. When you clip the wings and not
re-normalize, when you integrate, you don't get one.
MR. LONG: What I am saying is, when I
reintegrate, I effectively get one the way I did it. So I
wasn't worried about that part.
MR. HOLAHAN: You effectively add additional cases
to cover for the ones thrown away. It comes out the same.
MR. POWERS: Actually it's a nice analytic formula
for clipped wings where Gaussian distribution is not all
that hard to use.
MR. LONG: Considering that while I was doing this
the Sun station somehow changed their link to the subroutine
that gives me double precision random numbers to the point
that I realized that something wasn't right and I found my
random numbers were coming up between 0.4 and 1.8 and had to
go back and get a Fortran instead of a C subroutine, there
is a noticeable difference.
MR. POWERS: An absolute truism is never, ever,
never, never use a system's subroutine for any numbers.
Ever. There are no good ones.
MR. LONG: I did check them and I was getting a
curve that looked like I wanted it to look before I used it,
but I did that in 1996 when I wrote the program. Then when
I realized something was wrong in 1999, it came very late in
the process and it was kind of disruptive.
Trying to catch up on the schedule a little bit
here, I think this is all I want to say about uncertainties
right now. The next thing was the integrated decision
process, and I will talk a little bit more about
uncertainties in that if we are ready to go to it.
MR. POWERS: We are scheduled to take a recess
here for ten minutes or so.
MR. LONG: It sounds good to me.
MR. POWERS: Why don't we recess for 12 minutes.
[Recess.]
MR. POWERS: We will come back into session.
Next we will hear about the integrated decision
process.
What seems to have been badly misunderstood is we
had a contention on the integrated decision making process.
I felt an obligation to allow the staff to respond to any
contentions that they felt they would like to on the
integrated decision making. Looking through the viewgraphs,
I see that you really didn't choose to respond to the author
of the DPO but rather describe the integrated decision
making process. Looking through it, it looked extremely
interesting to me.
MR. LONG: He did not like the Farley decision. I
just wanted to describe the process with a couple of slides
so everybody is on the same page and then start talking
about Farley. That was the intention. So as I said, I will
talk about the five principles and then Farley and Arkansas.
The five principles that I still haven't learned
to recite in my sleep are, first of all, the proposed change
meets the current regulations unless it explicitly requests
some change, like an exemption;
The proposed change is consistent with the defense
in depth philosophy. Here we are talking about tubes that
are basically two of the physical barriers between the fuel
and the public. So that is an important one;
That it maintains sufficient safety margins. Here
we are talking about strength, leak rates, et cetera;
When a proposed change results in an increase in
core damage frequency or risk, the increase should be small
and consistent with the intent of the safety goal policy.
MR. POWERS: Do sufficient safety margins include
the time the operator has available to respond?
MR. LONG: Not explicitly in the sense that that
is not one of the safety margins that is in the design
basis. We talked about this. In trying to interpret what
defense in depth was, if your cut set comes to everything
works fine except you are relying on the operator, that is
not much defense in depth. So it comes down to how much of
the system do you really need to work right and how much
damage can one of those barriers give you, whether it's the
operator training or action, or whatever.
The impact of a proposed change should be
monitored using performance measuring strategies. Well,
some of these things are to take the steam generators out of
service and throw them away and put in new ones at the end
of the period of operation where they are requesting to not
do another inspection. It makes it kind of hard to figure
out if you were right about the degradation over the last
cycle.
Then there is consideration of uncertainties and
their potential effects on the decision, which I will try to
touch on again.
If we are clear on the principles, let's just dive
into Farley.
MR. POWERS: Let me make sure I understand. On
this plus point, this is not a requirement? This is
guidance to people who would care to make an application
under the guise of a risk-informed change to the licensing
basis?
MR. LONG: When you say a requirement, Reg Guide
1.174 is guidance, not a requirement. The whole process is
voluntary. But in it there are these five principles and
then there are some things that the guidance says they
should address, including the uncertainties.
MR. POWERS: If I came to you with an application
in which I had not considered uncertainties or their
potential effects on the decision and made a persuasive case
on why that was reasonable, staff would give it the due
consideration it deserved, right?
MR. LONG: I would always give it the due
consideration it deserves as soon as we have time. For
instance, I mentioned earlier South Texas has an application
in. They have tried to argue the tube support plates will
not move from the degraded portions of the tubes by more
than 0.15 inches, and they made the statement that they can
show that the probability of rupturing a flaw is 10 to the
minus 14th, assuming they have a flaw under every one of the
tube support plate intersections and that they all get
exposed by 0.15 inches. The way they did this was to take
the 0.15 inch length on the rupture correlation and figure
out how many sigmas there were to get down to the steam line
break pressure and then figure the probability of getting
there. I believe it was 10 to the minus 20. And then put
in something like 47,000 intersections that all had that
probability, and wallah, 10 to the minus 14. So the comment
back was we didn't think they properly considered the
uncertainties which were really controlling from the support
plate deflection calculation, and for that matter, the
ability to detect the flaws going beyond the support plate.
So, yes, we do look at the way they do things. We
get some amazing stuff in applications.
MR. BALLINGER: How much more amazing than that?
MR. POWERS: We got 10 to the minus 45th
probability of welds failing in the BWR. These guys aren't
even in the plausible lead right now.
MR. LONG: They did not consider the half life of
protons.
For Farley, this is the first time we tried to
apply this to the steam generator tube degradation issue.
As I mentioned earlier, Farley really didn't address the
principles in the reg guide, so I tried to go through and
elicit information with questions and write up an SER that
would be more like the guidance we never got out to the
licensee.
Based on their projection of the condition of
their tubes at the end of the cycle, they are projecting a
99 plus probability of withstanding design-basis accidents
of tube rupture and, I think, steam line break pressure
differentials. They were projecting a 90 percent
probability of withstanding severe accidents, which we kind
of agreed with in the calculation.
MR. KRESS: What does that mean, withstanding?
MR. LONG: Not having a thermally induced tube
rupture. Pressure induced didn't matter much here if you
believe the first bullet.
The condition of the tubes was projected to have
about a 50 percent probability of meeting the three times
normal operating pressure delta P. Deterministic process
would normally require 95 percent, and that is really the
reason they were putting in the application.
I did the projected LERF, as I have tried to
describe, putting in the uncertainties to get out one number
as opposed to trying to get out a distribution and figure
out what to do with the distribution against the numerical
guidance in 1.174, and it met the guidance by about a factor
of two. It's not a big factor.
MR. HIGGINS: Which sequences did you consider for
that, all those different types?
MR. LONG: Primarily, at this point I considered
the one that remained at normal operating pressure. We had
discounted the LOCA sequence on the basis of their seals and
some other thermal hydraulic changes that Charlie had made.
MR. HIGGINS: Was it just the high/dry ones, or
was it the normal spontaneous tube rupture or the thermally
induced one?
MR. LONG: Which question are you asking me about
here?
MR. HIGGINS: Number four.
MR. LONG: The primary contribution to number four
was from the thermally induced ruptures, because what they
were projecting was degradation that really would not be
susceptible to anything else with much probability. They
were 99 point something probability of not having a
degradation sufficient to produce a spontaneous tube
rupture. If you put that into the equation, it doesn't
affect the answer.
As I point out, the impacts were not monitorable
in this case because they were going to discard the
generators after the operating cycle, without inspection.
However, the way the tech specs work right now, they are
fairly weak because they are designed for the wastage. What
we did do was to use the Reg Guide 1.174 rubric to say if
they sustained some sort of leakage or other effect on the
steam generator tubes that indicate the degradation is not
as projected by them to get this license change, then in
accordance with the principles here they should go back and
do the inspection necessary to return it to that condition.
So we added a little bit of tooth to the amendment that way.
MR. CATTON: What happened to four if I went to
two and said that was 50 percent?
MR. LONG: Fifty percent?
MR. CATTON: In other words, I just flat don't
know which way it's going to go. Would that have still met
the 1.174?
MR. LONG: It probably would not have.
MR. POWERS: It depends a little bit on what you
define as a LERF.
MR. CATTON: I think 90 percent is too high.
MR. LONG: Too high to require or to too high to
believe?
MR. CATTON: Too high to believe.
MR. LONG: Can we go to uncertainties?
MR. CATTON: We just went through the
uncertainties associated with this. There is mixing; there
is the fact that the hot leg is treated as a tube. All
these things enter in. Where does it fall down? I don't
know.
MR. LONG: I will agree with you to the extent
that what I was doing here was going through a calculational
process as best I could at the time and coming up with
essentially 90 percent of the core damage. The high/dry
accidents were not resulting in bypass by the calculation.
In doing that calculation, I did take a look at the
variation of the temperature on the tube support sheet and
tried to integrate that in. I mentioned in the SER that
that was something I tried to do and that is something where
we needed more effort.
MR. CATTON: What is done is done. I was just
curious how much that probability of withstanding severe
accident would have to decrease before you don't meet 1.174.
If you can decrease it to 60 percent? I don't believe it's
90 percent.
MR. LONG: Just trying to remember where the
numbers came out, probably if you decreased it to something
like 80 or 75 percent you would be over the number for 1.174
small change.
MR. CATTON: So it's iffy.
MR. LONG: Yes.
MR. CATTON: I believe it's that number two that
is part of the DPO.
MR. LONG: That's correct. That is one thing that
Joe Hopenfeld doesn't think is correct.
MR. CATTON: He questions that mixing and he
questions the mixing probably because he sat in on some of
the subcommittee meetings that took place a few years ago.
MR. LONG: I have to agree that I have a problem
with the mixing as well. Remember, this is supposed to be a
risk-informed, not a risk-based process. The way I
approached this decision was not to say I know exactly what
is going to go on there.
MR. CATTON: I understand.
MR. LONG: The way the stuff that we know fits
together now with the logic we have this would look okay if,
and I will get to some of the uncertainties.
MR. KRESS: In the Reg Guide 1.174 risk acceptance
values, I think there is an implication in them that this is
a permanent change that is going to last for the rest of the
life of that particular plant.
MR. LONG: That's true too.
MR. KRESS: Here you have a temporary change that
is going to last a short time, which tells me you ought to
be able to relax the acceptance criteria by some equivalent
factor. Did that enter your thinking at all?
MR. LONG: Not by a particular factor, but it made
me feel a lot more comfortable about doing this.
MR. KRESS: It made you feel better about it.
Okay.
MR. LONG: I could only be wrong for a short
period.
[Laughter.]
MR. KRESS: If its remaining lifetime was ten
years and this was only for two years, I would have taken
the ten over two and multiplied it times that LERF and said
I could increase that acceptance value by that much. Or
something along those lines.
MR. LONG: We are sort of getting into the
philosophy of regulation here, but I think part of it is
what level of benefit you are getting and what level of risk
you are taking to get it. We don't really trade it off that
way explicitly, but in previous lives, dealing with other
logical decisions, there was sort of a rate of risk and rate
of benefit that you had to balance.
MR. HIGGINS: Doesn't Reg Guide 177 bring that
into time?
MR. KRESS: Yes. In fact, that is sort of what I
would have used, the time factor that they use in 177.
MR. POWERS: The problem is there is no delta CDF
here.
MR. HIGGINS: No, but there are delta LERFs.
MR. HOLAHAN: You can't do the same thing a decade
lower?
MR. KRESS: Anyway, I think the time at risk is a
consideration one ought to have.
MR. POWERS: I guess my feeling is when you are
talking about a cycle on a plant, 18 months or something
like that, I think you've gotten all the time you can get
out of me.
MR. HOLAHAN: I agree. We have in the past, and I
think maybe some of Steve's other examples have time as a
factor. Didn't we put time in Arkansas as a factor? But
when you get longer than one cycle, that is too long for me
too.
MR. LONG: I guess I should say that the delta
LERF was factored in in the sense that if it was for a
fraction of a year, we annualized it to a year.
MR. KRESS: Yes, you usually do that.
MR. LONG: I like to think of this as more a delta
in probability over the cycle or over a year. There are
other people who don't like to do their math that way, and
we get into arguments, but to me it always seems strange to
talk about the frequency of something that is only going to
happen once.
In looking at the uncertainties, I talked about
what I tried to do with the thermal hydraulic uncertainties
and I basically tried to give a lot of credit to the idea
that Charlie was nearly right and put some wings on it that
went out 50 degrees in each direction and integrate that in
the Monte Carlo process. There are the uncertainties in the
mechanical properties of the materials and so on that we
also used in NUREG-1570, so I won't go into that. They
weren't that important.
The biggest problem was the flaw size projections.
The reason Farley was asking for the license change was that
they had had a missed signal that turned out on the next
inspection to be a fairly significant flaw. Now they were
projecting to have corrected their inspection problem and
have a much better process and nothing like what they had
found last time should show up by the end of the next cycle
even though they weren't going to look.
What I did was a sensitivity study where I took
their previously found flaw distribution and put it into the
same calculation and came to the conclusion that that would
not satisfy Reg Guide 1.174. So that put it into the
materials people's lap to try to determine if they really
thought the inspection process had improved enough to grant
this license amendment. Based on what is documented in the
SER, they reached the conclusion that Farley probably had
been able to do that, and we granted the amendment.
We acknowledged the uncertainty for the 0.25 inch
crack length that was a threshold for cutting, and I won't
go into too much detail because I described that earlier.
This was the application that pointed out to us that we had
to deal not just with total crack lengths but with
through-wall segments of larger cracks in the Westinghouse
process for looking at the significant segment of a crack
that would either pop through wall or lead to a burst in the
weakest segment.
I think that is all I want to say about Farley and
will go into ANO-2, unless you have some more questions on
Farley.
The ANO-2 application was ultimately denied
earlier this summer. The reason really had to do again with
the NDE uncertainty. ANO-2 had a history of doing
inspections finding either just barely met the three delta P
or just barely did not meet the three delta P criterion,
shortening their cycle a little bit and running and doing
the inspection and finding essentially the same thing. They
were hanging in there around the 4,000, 4,300, 4,400 psi
pressure capability, and they were projecting that they
would be able to at least do that again if not better.
On the other hand, they were missing flaws that
were fairly deep and sizable in length, and we couldn't
reconcile their projection with what they kept finding, nor
could we find any plausible reason to believe that their
inspection had improved with the minor methodology changes
they had made. So we got into a problem of projecting
exactly what we should put into the calculation. That was
one of two significant problems we had.
Ultimately we ended up asking them to back
calculate their probability of detection and tell us based
on their previous two inspections what they thought the
probability of detection was as a function of flaw size,
which in their case was essentially depth; they didn't
include length in the detectability.
We found that flaws that looked like they would be
able to actually potentially challenge the main steam line
break criterion were flaws that they would not apparently
have a good probability of detection for. That is really
the basis for the denial.
One of the things that we came up with is this
last line here when we started looking at the uncertainty of
the strength of a flaw as characterized by NDE. They had
some full tube data where they had actually burst the tube
at a particular pressure, and they had characterized the
burst pressure as a function of a +Point profile.
What we really found was to get 95 percent
confidence -- and I don't know why I've got 5 there -- let's
put it this way. If you projected the flaw to have about a
4,000 psi strength, to get a 5 percent failure probability
you'd have to go all the way down to 2,700 psi. That is
quite a big difference. That's 1,300 psi. I think that was
the first time we realized how uncertain in terms of
strength the NDE characterization is.
We also had problems in the severe accident
calculational process. This licensee didn't come in and say
that the tubes would just not fail; they came in and said
the tubes will almost always fail. It's a CE plant. The
way they calculated it and the way we calculated it seems to
have a higher thermal challenge to the tubes. We're not
quite sure why. We know that the tube sheet is closer to
the top of the hot leg, that the plenum is not as deep, and
of course the tube sheet is broader. So there is a geometry
difference there.
Also, there is a higher power than we were looking
at in Surry.
So there are a lot of things that would tend to
heat faster, and of course, if you are heating fast, the
thin things tend to keep up with the gas. The thicker
things don't, like the surge line. So there is more of a
high temperature challenge in this particular plant.
The licensee was arguing that they were very
likely to fail tubes with flaws that were 30 or so percent
through wall and therefore there wouldn't be any delta LERF
if they had any larger flaws.
[Laughter.]
MR. BALLINGER: Run that by me again.
MR. LONG: What they were saying is that the flaws
that you could expect to be present in tubes and therefore
would probably even be in the hottest part of the plume
would probably fail under their high/dry sequences if they
depressurized the secondary side, and therefore having large
flaws really wouldn't change the outcome. So the delta LERF
was not there.
They had some other sequences that weren't quite
so challenging and they had some small delta LERF
contributions from those.
Another thing I want to point out to you is it's
very hard to go through these things and claim that you have
done them in a conservative manner, because then a licensee
will come in and turn the whole thing on its ear and
everything that you just did that was conservative is now
non-conservative. If you are going to do this business, you
can't just run everything off to the maximum on one side or
your delta LERF goes to zero on either side.
Another thing that happened in this calculation
was that they had looked at some intermediate pressure
sequences. I mentioned earlier they did that by setting the
set point down to 1,400 psi early in the transient and they
ended up with some fairly benign situations for those as
well. It lowers the delta P across the steam generator
tubes.
When we tried to duplicate those, we had some
problems, and frankly, at the moment I don't remember what
they are, so I won't go into it. That is what provoked us
to stick the pressurizer valves open by small amounts rather
than full open and got us into the type of thing that I
described. I showed you one of those stair-step creep
damage accumulations for a variety of different tube
strengths earlier.
What that turned out to be for us was essentially
calculational overload. Instead of being able to bring the
process to pinch points and talk about a small number of
options of where the flaw might be and what the temperature
might be there, we were looking at a very large number of
potential flaws that could be affected, depending on what
the temperature was. We really needed to do a volume
integral of everything all at once, and I did not have a
calculational tool developed that would just go do that.
We weren't ready to concede to them that
everything would simply fail. It looked to us as though
this is one of the cases where MAAP turned out to be more
pessimistic than RELAP. We don't normally see those, but
this was one.
MR. CATTON: It must have slipped by Bob Henry.
MR. POWERS: I'm stunned that he doesn't see those
more often. Usually when they find one of those, you can't
get away from it. They trumpet it in front of you all over
the place.
MR. LONG: Anyway, what ultimately happened here
was that we really figured that we could not deal with the
high/dry sequences for this case. We just couldn't
ascertain if we thought the delta LERF was low because too
many of them would fail, low because not many of them would
fail, or high because it was right on the edge of the cliff.
Their option for resolving that was to adopt a strategy for
depressurizing the RCS.
I didn't bring the graph with me, unfortunately.
They adopted a procedure and they made a plant modification
to allow them to carry it out. Because the high/dry
sequences were dominated by a loss of one dc bus and they
had not two RVs like most plants, but they had a path from
the pressurizer to the relief tank, it was blocked by two dc
MOVs, one from each safety bus. They needed to get them
both open to depressurize. What they had to do was come up
with a way of whichever bus was not powered get power from
the other bus to open the valve that was on the dead bus.
They put in a procedure. They put in essentially
some very large extension cords and proposed to us that they
would instruct the operators to depressurize at effectively
700k or 800 Fahrenheit. We asked them to tell us how long
they would have to wait after that period of time in order
to have at least a 0.25 or lower human error probability for
actually succeeding in taking the action to depressurize
given that they might have to go out of the control room and
hook up the extension cords.
They came back with a time frame that was a delay
of like 20 to 30 minutes. When we ran the thermal hydraulic
calculations, we essentially looked for the indication they
were going to have an operator assigned to stand there and
watch for the indication on the thermocouples. We waited 30
minutes and 27 minutes, and RELAP opened the valve. We
found that it looked very capable of depressurizing rapidly
enough to preserve the flawed tubes.
At this point we had put into RELAP the ability to
look at stress magnification factors, and we put in
magnification factors up to 7-1/2, I think, which is almost
ready to fail at normal operating conditions. In a creep
damage sense, that did not look like you came close to
failing those tubes even if you depressurized the secondary
side of the generator. I think at the end we had actually
melted the surge line and still hadn't brought the tubes
that would come close to the three delta P to failure.
So that looked like a successful process to us,
and that looked like they were on the way to some sort of
approval except they had the problem with the main steam
line break type of accident and not being really able to
demonstrate they could find the flaws that would threaten
during that accident.
I think that as far as I went on that one as well.
Are there any more questions on Arkansas or on the
integrated decision process or how it relates to the DPO?
MR. POWERS: I don't think so.
MR. HOLAHAN: From what Steve has done and from
the complexity of the earlier discussions it is pretty clear
that if we are doing anything as difficult as these cases,
we are not going to do a generic analysis and say, yes, our
generic insight is that this sequence is important and this
one isn't. They are far too plant specific, and in fact
they turn out to be often cycle specific, because you have
to have pretty good insights as to the latest inspection
information so that you have good information on the flaw
distributions and things like that.
So even though we feel good about having increased
our capability of dealing with these sorts of issues, they
are very difficult to do, they are very time consuming, they
are very plant specific, and one hopes not to have to do
this sort of analysis often. We would prefer to have steam
generators with fewer flaws and maybe not quite being pushed
so hard.
MR. LONG: I hope I have conveyed some of my
discomfort level in trying to do these things. They really
are in an extremely uncertain area and it's difficult to say
that you have done something like this in a defensible way.
In terms of risk informing something, I think you can
honestly say if we are doing it today, this is our best
guess at what the answer is in risk space, but I don't think
we are ready to come close to being risk based in this
particular area.
The other thing I would like to acknowledge and
comment on is you notice there was a backfit here to
depressurization to avoid the LERF component from a high/dry
sequence. That was the thing that we supposedly did a
generic backfit analysis on back in the days of rulemaking
and decided that, gee, we couldn't see a backfit that would
be justifiable on the basis of the LERF component from
high/dry even if all high/dries are LERF. I think that sort
of calls that conclusion into question a little bit, because
it really had to go to, well, how much does it cost to make
the change that might be beneficial. I think we have here
an inkling that it's not too difficult to be pretty
beneficial.
MR. STROSNIDER: One other comment with regard to
these two plant-specific amendments to make sure it is clear
to everybody. The degradation that was of concern was not
involved with Generic Letter 95-05. It was other forms of
degradation that was driving these analyses.
MR. POWERS: We come now to the section of the
agenda that involves a summary. Before we get into that
summary, I will relate just a little bit of an anecdote to
people.
As you might have suspected, I have spent the week
having people sidle up to me and saying, how is the DPO
stuff going?
[Laughter.]
MR. POWERS: And I have given them a very positive
response. I said to them I think it's going extremely well,
and I think the reason it's going extremely well is we are
getting outstanding presentations from the NRC staff and got
an outstanding presentation from the DPO author. Since I
notice not all managers but several managers are here, I
hope you will pass on to your staff and, if you have the
opportunity, the DPO author that I think you guys have done
a bangup job presenting this material. It's just an
outstanding job, and I think I have gotten that same sense
from my entire committee.
MR. STROSNIDER: Thank you. I do appreciate those
comments. We will feed it back to the staff.
I noted in my introductory comments you were going
to hear from a wide variety of disciplines and people. I
think that indeed we do have a very dedicated and
professional staff.
MR. POWERS: I think you should be very pleased at
the way they have been able to work together on these. We
see an unexpected amount of coordination between the
disparate disciplines.
MR. STROSNIDER: It's a real statement about the
movement towards risk informed. We have got metallurgists
asking questions about LERF and CDF, and we have got risk
assessment people coming down and asking about metallurgy,
and it has been very beneficial.
MR. POWERS: The difference is the metallurgists
get answers.
MR. BALLINGER: Both are black arts.
[Laughter.]
MR. STROSNIDER: This suggests that I am going to
give a summary of steam generator issues. I'm not going to
summarize the last three days.
MR. POWERS: I thought you were going to write our
report for us.
MR. STROSNIDER: I would like to make some brief
conclusionary statements and maybe just touch on a few
thoughts that I hope people will carry away from this
meeting.
First, I want to emphasize that the staff does
take the DPO issues and steam generator issues very
seriously. When I put this viewgraph together it was
intentional in the title there where I said "DPO/Steam
Generator Issues." You heard a lot of stuff in the last
three days. Some of it is directly related to the DPO and
intended to address that, and as I said in the introduction,
some of it goes beyond the DPO.
There has been and remains something of a
challenge of identifying exactly what is in the DPO and what
other issues the staff may have taken on as a result of some
of the rulemaking exercises and our improved understanding
from a risk-informed perspective and trying to move that
forward.
Regardless of whether they are DPO issues or other
issues that the staff is pursuing, we do take them
seriously. A couple examples here.
There is an extensive amount of documentation on
these issues. I think somebody said 89 pounds.
MR. BALLINGER: To be exact.
MR. POWERS: And it is going up.
MR. STROSNIDER: There has been a lot of thought
and a lot of work that has gone into this. I will come back
again to the offer and before I do finish we will talk a
little bit about future coordination. Where the staff can
be of assistance in helping to point to the right reference
and the right section of a reference to help answer any
questions you've got, please let us know.
Development of regulatory framework. I didn't
plan on getting into a lot of detail on this, but as we move
forward in the new framework that is being developed with
the NEI 97-06 guidelines and tech spec change framework we
are taking these things into consideration.
I gave a few examples the other day where, for
example, the industry wanted tech specs that would allow
them to establish repair criteria, that would allow them to
establish repair methods. We said, no, you need to bring
those into NRC for review and approval. The reason for that
is we want to make sure that we can look at the kind of
issues we've been talking about.
With regard to the plant-specific evaluations,
Steve just went through two of those. I think the main
point I wanted to make there is that the staff, number one,
said that we were going to consider these things. Some of
the resolution of the way we are addressing the DPO issues
is we said we are going to consider them in our process.
As I said the other day, we never know what the
next alternate repair method or the next risk-informed
amendment is going to have in it. We did consider them, and
we can get into some discussions about do we have the best
models, can we improve on them.
The answer is, yes, we can improve on them. But
we did consider them, and I think we demonstrated that the
NRC staff and NRC management is willing to make some tough
decisions following these guidelines. When we denied the
Arkansas request, they shut down for something like two
months before the scheduled steam generator replacement
outage to perform a steam generator inspection. That is not
a decision that can be made lightly. It wasn't, but we put
it through this process. We considered the risk insights
and we made that decision.
With regard to research activities, I think you
have seen through the last couple of days that we have also
had very close cooperation between NRR and the Office of
Research. Where we see that we need to make improvements,
where we can improve in our models and where we want to do
that in order to apply it in the licensing process we are
asking them for assistance. We talked about the tube
cutting, and they came back with some very good information
this past week to address that issue. We have asked them
now to look at the vibration issue, and they are doing that.
We all have the users meet, which covers a broader
spectrum of risk-informed issues, ranging from the thermal
hydraulics to some of the tube failure, the surge line
response, and whether the creep modeling there is as good as
it should be.
Where these issues come up we are taking them
seriously; we are pushing toward resolution on them. When I
say resolution there, I guess maybe the thing to say is
improve our understanding in some of these areas.
Maintaining safety. During Ken Karwoski's
presentation he put up a viewgraph demonstrating that the
number of tube leaks and forced outages has decreased,
depending on when you start looking at that. If you go back
into the 1970s or early 1980s, there has been significant
reduction. But I think we do need to give some credit to
the industry, and I think the NRC staff has also had some
influence on that. People are applying improved
technologies today and I think there is some benefit there.
Risk-informed approach. As you can see, we are
moving into that. We have applied it now in several
licensing actions. I think those insights are helping us to
maintain safety. This last example where Arkansas
identified what they could do to reduce the frequency of
high/dry events and help out with that is a good example of
how this is helping to maintain safety.
I think everybody hopefully has heard and is aware
that we have four management goals. You can find them in
our strategic plan: maintaining safety, reducing
unnecessary burden, improving public confidence, and
improving efficiency and effectiveness in the realism of our
decisions.
Given those four outcomes, maintaining safety is
the priority. I hope that when people see the approach that
we are taking that they will appreciate that that is our
perspective.
Future actions. Shortly after the Indian Point 2
tube rupture on February 15 the NRC initiated a lessons
learned task force. It's another multidiscipline,
multi-office effort. We expect to see the results from that
report shortly.
We also have a report that was done by the Office
of Investigation which has some observations in there.
We are taking those reports and we will be looking
at where we can improve our processes internally, and we
will also be looking at what areas we need to address with
the industry in terms of improvements that can be made.
With regard to the NEI 97-06 license change
package, that was put on hold after the Indian Point 2
event. That was a conscious decision that we didn't want to
go forward with approving that framework methodology until
we understood the lessons learned and could factor those
into our review.
MR. POWERS: One of the issues that I have been
wrestling with is whether to try to factor in the Indian
Point 2 event and the lessons learned into this DPO
resolution process. I understood you all had been resisting
doing that, because at 89 pounds one more piece of paper did
not seem to be an absolutely essential thing, but I would
appreciate your perspective on whether we should or should
not be looking at the event and anything that comes out on
the lessons learned.
MR. STROSNIDER: It comes back to the comment I
made earlier, which is that it has been somewhat of a
challenge to define the scope of the DPO. When you look at
the root cause and when you look at what we are pursuing in
this area, I would point to the inspection report and the
proposed enforcement action that is under consideration now.
It has to do primarily with licensees following Appendix B,
the quality of their program, actions they could have taken
with regard to improving the quality of the data, following
up after they found an indication that was similar to the
one that failed, and some actions like that.
I don't see that those were areas that were
addressed in the DPO. There are some broad issues in the
DPO about probability of detection and eddy current testing.
When you go back and look at a lot of that which was raised
in the context of initially the voltage-based approach, I
don't see that it was something directly raised in the DPO.
Whether you want to take a look at what is going
on there to inform just what is going on with steam
generators in general, that is another question. As I said,
we will be coming out with reports in that area in the near
future.
MR. HIGGINS: Is there anything significant that
is coming out generically from Indian Point, or is it mostly
plant-specific items?
MR. STROSNIDER: The industry is doing a lessons
learned effort on this as well as the NRC. One of the
things clearly that is being looked at is generic
implications. If you go back and look at this failure, one
of the main contributing causes was poor quality of the eddy
current data. They missed a very large indication. A
hindsight review, knowing where it failed, they went back
and looked at the data that had been taken in the inspection
prior to the failure. They were able to see this
indication. They went to some higher frequency eddy current
data that made it easier to see.
There are some techniques that can be used to
enhance that, using higher frequency eddy current. Some
plants have already gone to those higher frequency probes
and doing the U-bend inspections. We had a meeting with the
Nuclear Energy Institute and the Electric Power Research
Institute. They are going back and modifying the EPRI
guidelines on qualification, and they are going to address
this data quality issue.
So, yes, there are some generic implications.
During this outage season when the staff is talking to
licensees that are doing inspections we are asking them what
they have done to address the lessons learned from Indian
Point.
With regard to 97-06, we will be re-initiating
that review in the near future. I think originally we were
scheduled to come talk to the ACRS about that. I think it
was at the December meeting. Don't hold me to that. It has
been rescheduled. Given the delay that we consciously took
with regard to this review, it is going to be more like the
March time frame, but we will be back talking about that
framework. We still think this is a good thing to pursue.
MR. POWERS: I guarantee you that if ACRS had the
opportunity to delay anything out of December, they did.
MR. STROSNIDER: I mentioned that the PWR
licensees have committed to follow guidelines. In fact,
they have done some update on their own to reflect some new
improvements. When we start talking about these condition
monitoring and operational assessment type things, this is
the reason that licensees are doing it. So clearly there
are some improvements here.
I guess the final thing is again I want to thank
all the committee members here. This is a tough area. We
appreciate the time and energy that you are taking to look
at it. It's an opportunity for the staff to hopefully see
some resolution to some of these issues, and we clearly want
to support that. Whatever we can do to help in your
deliberations, please let us know. I guess the process for
doing that would be Undine could contact me.
I'm afraid this probably isn't a comprehensive
list, but I did put together some of the things I noted
during the discussions that I think we owe you now. I can
run through that briefly if you would like to hear that.
The first item is to provide some more information
on how the Generic Letter 95-05 leakage values were adjusted
for pressure and temperature.
The second item is some additional discussion on
the basis for the 10 to the minus 2nd conditional
probability of tube failures given a main steam line break.
That is the criteria that is in Generic Letter 95-05.
We were asked to provide the distributions used in
Generic Letter 95-05 for analyst variability and probe ware.
We will provide those.
We were asked in the proprietary information you
have that shows the data points associated with the leakage
and the burst correlations to identify which of those data
points are from tubes that were pulled from the steam
generators versus tubes that were manufactured in the
autoclaves in the laboratories.
You wanted to see the information regarding the
Maine Yankee circumferential cracking. We will provide some
of the metallography and the pressure test data that show
how those type of cracks respond to that type of load.
We are going to see what kind of information we
can gather with regard to the Turkey Point event that Mr.
Spence discussed. Specifically, we want to find out if
there was post-event inspection done and what the results of
that inspection were.
MR. POWERS: Any evidence of permanent
deformations and things like that would be especially
interesting.
MR. STROSNIDER: Frankly, I think there must have
been some inspection done after that before its generators
were declared ready to go into service. It's just a matter
of seeing if we can find some of the documentation.
MR. POWERS: It may not be very extensively
documented. That is the headache you have.
DR. SIEBER: It was 30 years ago.
MR. CATTON: It was 1973, wasn't it?
DR. SIEBER: 1971.
MR. STROSNIDER: We are going to see what we can
find.
Dr. Ballinger was talking to us at the break about
providing some clarification on some assumptions that were
made in the Indian Point 2 significant determination
evaluation, specifically with regard to some of the human
reliability assumptions. We will get back with that
information.
That is the list that I had.
Steve, there was some discussion where I think you
were committing to provide some information. I didn't get
that written down.
MR. LONG: There are two more things I have on the
list.
One is the consequence difference for having 100
gpm primary to secondary leakage sized hole. It doesn't
change through a core melt accident that eventually fails
the RCS and the containment. That was the work that
Research had done. I was trying to guess what the
consequence relationship was to a contained accident. We
will get you that.
Also, I had made reference to some work the French
had done, which I think is proprietary, trying to look at
the effect of the crud in the crevice and the drill hole
support plate. We will get you something about that.
MR. HOLAHAN: I think we talked earlier about
providing some additional information on the iodine spiking.
To the extent we can address some of the questions that were
raised, for example, which data points have
depressurizations in them, and sort out some of those
issues, we will pull that together as well.
MR. POWERS: Let me share with you what I think
our schedule is going to be, with a great deal of
tentativeness, because, quite frankly, we won't know for
sure until next week. Our intention is to try to put
together a draft report from the panel over the remainder of
this month, which may have holes in it, but enough so that
we can pass it on to the peer reviewers we have identified,
who are members of the ACRS, by and large, and present it to
them at our November meeting. At that November meeting we
will give them some sort of a synopsis of what we have done,
kind of a status report of where we are.
I am allowing in that November meeting time for
the DPO author and the staff to make any rebuttals to things
that they have heard about. I've been told the DPO author
wants to come speak. It's not a great deal of time. We are
looking for fairly succinct operations. It's going to be
about half an hour for each side. If you care to make any
additional comments at that time, there is a block of time
available there.
MR. HOLAHAN: Do you know what day that would be?
MR. DURAISWAMY: It's November 2 at 2:30 and 4:30.
MR. POWERS: The peer reviewers on the ACRS would
have about two or two and a half weeks in November to
prepare their comments and get them back to us. The panel
will try to revise its report to accommodate their comments
so that we can provide a final report and maybe even a draft
position paper for consideration by the ACRS at our December
meeting.
Again, I suspect that we will allow time for any
additional comments at that time, but it will be relatively
brief periods of time. Just the exigencies of the FACA, it
seems that I have to allow time in there, but I haven't
figured out exactly what it is.
In other words, I would hope that we would provide
such a sterling report to the ACRS that they could move
forward promptly to provide an approval letter that they
could send to the EDO. It is my hope that we can wrap up
our portion of it no later than the middle of December. I
have no idea what schedule the EDO would operate on from
there. It's kind of his bailiwick. I'm moving on a pathway
for a prompt resolution on this right now.
This can be upset if in our discussions tomorrow
we find out that there is some glaring hole, but quite
frankly, I haven't seen any glaring hole. I think these
things have been very complete, very thorough, and very well
presented so that we understand where everybody stands on
all of the pertinent issues. I'm optimistic of meeting my
schedule.
MR. STROSNIDER: Thank you. This helps us do our
scheduling and gives us some idea of how quickly we ought to
be getting you some of this information, which we will do as
quickly as we can.
MR. POWERS: I think we will probably be making
changes in this report to the ACRS right up until December
1. Once the report gets to the ACRS, changing it after that
becomes troublesome to me, aside from editorial and cosmetic
changes. The actual report to the EDO that they make, of
course, is up to the ACRS. I have truthfully no control
over them, especially the distinguished representative from
Tennessee.
MR. KRESS: That's right. I've been known to
throw bombs.
[Laughter.]
DR. SIEBER: Did the peer reviewers get all the
documents that we got?
MR. POWERS: I think they have access to all the
documents.
MS. SHOOP: They got the first box. They didn't
get the second box that we gave to you guys today.
MR. DURAISWAMY: We'll send it to them.
MR. POWERS: Our peer reviewers are by discipline.
I'm not asking them to peer review all the documents save
what they want to say about them.
MR. STROSNIDER: In your discussions tomorrow, if
you come up with additional information or requests, we will
get those from Undine and we will respond to those.
MR. POWERS: Undine will still function as the
point of contact between the panel and everybody else
involved.
MR. STROSNIDER: Once again I do appreciate and
want to express appreciation on behalf of the staff for your
efforts in looking at this issue. It takes a lot of time
and energy, and we appreciate your help in addressing the
issues. Thank you.
MR. POWERS: Thank you.
At this point what I want to do is turn to our
consultants and ask if they have any comments at this stage
that they would like to make orally on what they have heard.
We do ask that you provide us a written report. Anything
you would like to pass on to us at this point would be
appreciated.
MR. CATTON: Me first?
MR. POWERS: Why not, you being the shy and
retiring type. We've got to draw Ivan out a little more.
MR. CATTON: That's right.
I think it has been an interesting exercise,
particularly tracking through the sequence of reports
written by Joe Hopenfeld. He really got better and better
at writing them as he went along. I was perplexed by the
staff's responses, because they didn't seem to change very
much. But the last three days I think they have done a very
good job. I think the response is here. The question is
whether or not you like the response.
The one area is the heatup during severe
accidents. I think that is fraught with uncertainty and I
have felt that for years. I just keep saying the same
thing, but there has not been much response.
Mixing is an issue. Heat transfer from one end to
the other is uncertain, and what do you do with it? It
seems to me you ought to assume a 50 percent probability of
failure of one over the other and be done with it, or you
have got to spend a lot of money.
The other area is the response of the system to a
steam line break, the whole blowdown process, what happens
inside. This is not a new issue. This was discussed in the
early 1970s, and one of the consultants to the Thermal
Hydraulic subcommittee even wrote reports on it. He tried
to retrieve them, but they are in Word Star, and Word Star
doesn't translate anymore.
I was impressed with what is done with the
relationship between voltage and burst and voltage and leak.
It seems to me that just bounding would put that to rest.
I don't see a lack of correlation like I heard. I
don't recall who was making the presentation, but they
argued for using a mean value. Leakage through these cracks
is just like flow in porous media. There are a whole lot of
parameters that are at the micro level and you are trying to
do something at the macro level, and your microscopic
variables are delta P and flow. Unless you incorporate the
variables at the bottom level into the equation, you are
never going to get it right.
In heat transfer we are faced with three or four
decades of variation for a same kind of problem. I think
you have got to choose the top or the bottom, depending on
whether you are buying or selling. In this case here it's
safety. You've got to choose whichever side of that band is
the worst. I think to put a mean through the curve is
inappropriate, but that's a personal view.
I like the process of going from the distributions
and how you extrapolate them all the way to either leakage
or burst. Some of the details in between probably could be
tightened up a bit.
I think making measurements on the pulled tubes is
going to help. The interesting thing is that you use the
voltages in situ and then you test the tubes after you pull
them. That can't make everything worse. So that puts a
conservatism in the ballgame, which I think is nice.
The other thing I was a little bit bothered with
is how you treat the iodine. I don't think there is any
question about the bottom line because there is so much
conservatism, but whenever you justify a poor model by
arguing conservatism somewhere else, I think you put a major
problem in front of yourself. You've got to deal with it.
You should take your best shot all the way through and then
add a safety factor if you are uncertain, not a huge
conservatism in one place to cover the bad modeling in
another.
All in all, I think it was pretty good. I think
the staff has really done a good job in coming to grips with
all of the issues.
MR. POWERS: Ivan, I think I and the other members
have a pretty good understanding of your concerns over the
mixing and heatup area. To the extent any written report
focuses an area, the more you could offer us on the
relationship between voltage and flow, I think that would
help me the most.
MR. CATTON: I don't know a whole lot about their
problem. I will put something in there. Whenever you are
addressing a problem that is some kind of transport
phenomenon in a heterogenous media, and particularly when
it's hierarchical from small scale to large scale, you have
a major headache in coming to grips with that kind of
problem. It is only in recent years where people are
actually developing the tools to do it.
MR. POWERS: I would appreciate comments that you
would like to make on the standards that you would expect
within your technical domain for that kind of a problem.
MR. CATTON: I'll do that.
MR. POWERS: Thank you.
MR. STROSNIDER: Dr. Powers, I don't want to get
into a whole lot of extended discussion, but we will provide
you some additional information. With regard to the
leakage, we talked about using the mean value. In fact, I
think what is used in 95-05 is a 95 percent confidence
value, but we will get you a clear description of that just
to make sure that there is no misunderstanding.
MR. CATTON: I took a look at one of those
figures. I don't know where I got them, but they got the
yellow sheet on them. And I get a really nice relationship:
LPH equals V.
MR. STROSNIDER: We will provide some more
information.
MR. CATTON: On the burst, I guess that was the
7/8 tube. On the 3/4 inch tube if I use 2 V, it works
really well.
MR. POWERS: Jim, I think I called you too
quickly. I need to ask the rest of the panel if they have
any questions of Professor Catton.
[No response.]
MR. HIGGINS: A general comment first. I thought
it was a worthwhile exercise that we all went through.
Looking at the stack of documents and what has happened over
the years, I feel that the DPO has clearly been around too
long and it's time for resolution and I think it's ripe for
resolution also.
I think the presentations that we got would allow
us to resolve most of the issues. There are clearly a few
things hanging out there that should be addressed either by
the staff or the industry. There are also some things that
have been indicated that are being worked now by Research or
NRR that need to be resolved but are under way through the
existing processes. I would support and I would hope the
rest of the committee here supports trying to resolve this
through the efforts that we are doing over the next month or
so and not just putting it off to some other committee.
I broke my comments up into two areas. One is
design basis and the other is severe accident, because I
think the DPO addresses both of those.
I think the design basis cases relate mostly to
Generic Letter 95-05 as far as the DPO lays it out, and
think that whole Generic Letter 95-05 process is very well
laid out and the analyses that are laid out there and the
bases and the background for them are good. It seems like
the submittals that are coming in are pretty reasonable too.
There were a lot of areas questioned by the DPO,
and without ticking them off, it seems like the staff made
convincing presentations on most all of those that what they
are doing is very reasonable.
A couple stick out as being questionable. One
that has been colloquially called the wild and wooly main
steam line break is one that is clearly an issue still, but
that looks like it's going to be treated by GSI-188.
Without having been able to read specifically what goes into
GSI-188, it seems like it may be constituted a little bit
narrowly to address all the concerns that were identified by
the DPO and that are probably legitimate concerns. It seems
like it may be limited to only the residents when there are
other displacement type of activities that have been brought
up.
The second area that seems to be open. I second
what Dr. Catton said about the iodine spiking. It seems
there is lacking a sufficient technical basis for the
calculation of the 335 factor and also for the 500 factor.
I don't doubt that there is plenty of conservatism in the
other areas that could account for that. Maybe that issue
together with the issue that we discussed quite a bit that
Dr. Bonaca brought up on the 30 minute for operator action
is a good reason to try to revisit the methodology for the
design-basis recalculation of design-basis steam generator
tube rupture analyses and to fix those.
On severe accidents, again it looked like the
staff has done a lot of work there and presented a fairly
convincing argument that most of the severe accidents
associated with steam generators have been reasonably
addressed.
It seems to me like there are three general types
of these. One is the thermally induced rupture after a core
damage event; one is the spontaneous steam generator tube
rupture; and the others are various transients that lead to
core damage from other initiators that result in abnormally
high DP's across the steam generator tubes.
It seems like they have all been reasonably
addressed, with a few comments, some of which pertain to how
you would look at severe accidents when considering Generic
Letter 95-05. The reason I bring that up is because the DPO
does bring up how you would, from a severe accident
standpoint, consider the things that are being done in
Generic Letter 95-05, and even though that may not have been
brought up at the time, it is certainly appropriate in the
current days regime of risk-informed regulation to look at
that.
It did not seem like the assumption of the
restraint of the Generic Letter 95-05 by the TSP was
adequately justified by the staff. It may be legitimate,
but I didn't see a good justification of it in the documents
or in the presentation.
Secondly, I heard also the staff say that they
believed that the CDF and LERF increases due to the Generic
Letter 95-05 exceptions were considered to be zero or small
but again did not see any quantitative presentations on
that. It seems like that is something that should be done,
and I'm not sure if it needs to be done generically or on a
plant-specific basis. Steve mentioned that it looks like it
is being done now on a plant-specific basis, and maybe, if
that is the case, you don't need to do it generically.
I would have liked to have seen some generic
presentation that considered the three different types of
severe core damage accidents associated with steam
generators. You can make an argument that all of those have
some potential of being affected by the 95-05 relaxation.
I guess I would also comment that the discussion
that we had on the HEPs and the concern that the DPO had, I
didn't really see a problem with respect to what has been
done over the various things. Even though that is an area
where there is considerable uncertainty and variability, I
felt what has been done in the various studies, especially
the more recent studies, is reasonable.
MR. BONACA: I'm sorry. Could you repeat that?
Reasonable regarding what issue?
MR. POWERS: Human error probabilities.
MR. BONACA: Okay.
MR. HIGGINS: The issue that they had raised on
the human error probabilities associated with the tube
rupture sequences.
There were a few other areas of the severe
accident that were raised by the DPO that have not been
addressed to date but are being addressed by the new
research-related areas that are going on as a result of that
February 8, 2000, letter.
That's all I had. Thank you very much.
MR. POWERS: To my mind all of your comments are
very useful. In your written report, I think it would
probably be of most use to the committee if you could focus
on what your thinking is about this 30-minute operator
action under your design-basis activities. Similarly, your
thoughts on the HEP, the more recent studies.
All the comments are good. If you have an
opportunity to focus, those are the two areas that I think I
could use the most help from you.
I will turn to the rest of the committee and see
if they have any suggestions.
MR. BONACA: Any comments on the HEP. Expressing
perspectives and opinions is still a soft area, particularly
when you get into multiple tube ruptures and so on and so
forth, which is really the area of concern presented by the
DPO. Any insights on that would be useful.
MR. HIGGINS: Okay.
MR. POWERS: We are going into a fairly intensive
activity tomorrow as a panel. I think we need to go back to
the contentions list and try to walk through those
individually tomorrow, deciding what we are going to write
and what we are going to say in something of an outline.
I will remind you that the author of the DPO
provided us a list of questions in addition to his
contentions, and I think I have to treat those under the
contention category that has been laid on the table and to
at least deal with them. If not as specific contentions,
our response should address those. So I will encourage you
to take a look at those questions to make sure we have
answers.
Finally, in the spirit of the point that Dr.
Catton made that the DPO author gets better and better at
articulating his point, he did conclude his study with two
recommendations, that Generic Letter 95-05 should be
withdrawn and those plants that use the alternate repair
criteria should be shut down. I had hoped we would not have
to address those, but I think we will have to give a very
clear recommendation in regard to both of those
recommendations that he has made. So be prepared to discuss
those as well as the more detailed technical contentions.
MR. CATTON: His first one was 95-05. What was
the second?
MR. POWERS: That those plants that have the
alternate repair criteria be shut down.
MR. CATTON: The 17 plants.
MR. POWERS: It's the ones that have the alternate
repair criteria, and I think in his oral presentation as
opposed to what he has written down those that don't go
immediately to the 40 percent plug-in criteria of old should
be shut down. I think we have to address that. I think we
have to give something very explicit in the report on that.
Do any of the members have comments they would
like to make, any comments on the overall strategy that we
would need to think about tonight? We will undoubtedly find
ourselves revising and honing this strategy a good deal
tomorrow.
My thought is that I will probably lose quorum
tomorrow about one o'clock. That is typically when I lose
quorum on these things.
We do have a little challenge getting into the
building tomorrow. We have to go by way of subterranean
passages.
I want to think you, Jim and Ivan. I think you
have added to this. I think your reports are going to add
to this. It is very helpful to have you here. I look
forward to what you have to say. As this draft report comes
along I will be sending you copies and looking for your
comments and any advice that you could offer to us. You can
switch hats and start playing the role of peer reviewer
here.
Any other comments?
[No response.]
MR. POWERS: Again, my sincere appreciation for
the quality of work by the staff and their presentations and
their managers. I think you've made this task a lot easier
than I forecasted it would be.
With that, I will recess, and that ends the need
for recording.
[Whereupon at 6:45 p.m. the meeting was
adjourned.]
Page Last Reviewed/Updated Tuesday, July 12, 2016