Thermal-Hydraulic Phenomena - November 28, 2001
Official Transcript of Proceedings
NUCLEAR REGULATORY COMMISSION
Title: Advisory Committee on Reactor Safeguards
Thermal-Hydraulic Phenomena Subcommittee
Docket Number: (not applicable)
Location: Rockville, Maryland
Date: Wednesday, November 28, 2001
Work Order No.: NRC-116 Pages 1-369
NEAL R. GROSS AND CO., INC.
Court Reporters and Transcribers
1323 Rhode Island Avenue, N.W.
Washington, D.C. 20005
(202) 234-4433 UNITED STATES OF AMERICA
NUCLEAR REGULATORY COMMISSION
+ + + + +
ADVISORY COMMITTEE ON REACTOR SAFEGUARDS
(ACRS)
THERMAL-HYDRAULIC PHENOMENA SUBCOMMITTEE
+ + + + +
WEDNESDAY,
NOVEMBER 28, 2001
+ + + + +
ROCKVILLE, MARYLAND
+ + + + +
The Subcommittee met at the Nuclear Regulatory
Commission, Two White Flint North, T2B1, 11545
Rockville Pike, at 8:30 a.m., Graham B. Wallis,
Chairman, presiding.
COMMITTEE MEMBERS:
GRAHAM B. WALLIS, Chairman
THOMAS S. KRESS
F. PETER FORD
GRAHAM M. LEITCH
WILLIAM J. SHACK
VIRGIL L. SCHROCK
PAUL A. BOEHNERT, Staff
RICHARD LOBEL, Staff ALSO PRESENT:
STEVE BAJOREK
HARV HANNEMAN
ROBERT HENRY
JOSEPH M. KELLY
NORM LAUBEN
JOHN MAHAFTY
JOE STAUDENMEIER
MIKE TESTA
JENNIFER L. UHLE
TOM ULLSES
WEIDONG WANG
A-G-E-N-D-A
Introduction . . . . . . . . . . . . . . . . . . . 4
NRC RES Presentation:
T/H Phenomena Research Program . . . . . . . 4
Assessment & Quantification
Steve Bajorek. . . . . . . . . . . . 155
Status of Experimental Programs
Steve Bajorek. . . . . . . . . . . . 183
Realistic Analyses of Large Dry Containment Response
to DBA Using EPRI MAAP Code
Robert Henry . . . . . . . . . . . . . . . 252
Mike Testa, Beaver Valley. . . . . . . . . 253
Harv Hanneman, Point Beach . . . . . . . . 258
NRR Presentation
Comment on Realistic Model Approach
Richard Lobel. . . . . . . . . . . . . . . 361
Adjourn. . . . . . . . . . . . . . . . . . . . . 369
P-R-O-C-E-E-D-I-N-G-S
(8:32 a.m.)
CHAIRMAN WALLIS: This is a meeting of the
ACRS Subcommittee on Thermal-Hydraulic Phenomena. I
am Graham Wallis, the Chairman of the Subcommittee.
Other ACRS members in attendance are Peter
Ford, Thomas Kress, Graham Leitch and William Shack.
ACRS consultant in attendance is Virgil
Schrock.
The purpose of this meeting is for the
Subcommittee, firstly, to continue review of the NRC
Office of Nuclear Regulatory Research Activities
pertaining to thermal-hydraulic phenomena in support
of the ACRS annual report for the Commission on the
NRC Safety Research Program.
And secondly, discuss a proposal by the
licensees of the Point Beach and Beaver Valley plant
to perform more realistic analysis of the containment
design basis accident EPRI/MAAP code.
The Subcommittee will gather information,
analyze relevant issues and facts and formulate
proposed positions and actions as appropriate for
deliberation by the full committee.
Paul Boehnert is the cognizant ACRS staff
engineer for this meeting.
The rules for participation in today's
meeting have been announced as part of the notice of
this meeting previously published in the Federal
Register, November 15, 2001.
Portions of the meeting will be closed to
the public as necessary to discuss information
considered proprietary to the electric power concerns.
A transcript of this meeting will be kept. And the
open portions of this transcript will be made
available, as stated in the Federal Register notice.
It is requested that speakers first
identify themselves and speak with sufficient clarity
and volume so that they can be readily heard.
We have received no written comments or
requests for time to make oral statements from the
public.
Now, our hope as a Subcommittee is that
today's meeting will be the highlight of the year as
we hear about all this great work which is going on.
I call upon Jack Rosenthal to get us started.
MR. ROSENTHAL: Thank you. I'm Jack
Rosenthal. I'm the branch chief of the Safety Margins
and Systems Analysis branch in the Office of Research,
and I just have some introductory remarks and then, as
you can see from your agenda, Jennifer will talk about
applications, Jennifer Uhle. And then Joe Kelly about
code consolidation and Steve Bajorek about our
experimental program, and I'll get help from much of
the other staff.
But I wanted to make some introductory
remarks in a few areas. The easiest one is that we've
accumulated a fair amount of hardware now, and so at
least the capability to run the codes. And we're
proud of a new PC cluster that we're doing in CFDR.
The next thing, more important, and I
don't want to embarrass my staff, but we have now in
fact I think a world premier staff of people that have
come on board, and several recent ones.
Joe Kelly was at the NRC and has returned.
Steve Bajorek was at Westinghouse in
Kansas State and is now with us.
Joe Staudenmeier and Tony Ullses were in
NRR and have joined us.
Chris Murray was at Penn State and has
joined us.
And so we have a staff that's now capable
of analyzing experiments, developing the codes and
doing the safety analysis. And we should be proud of
the staff.
In terms of products, okay, we are using
our codes to make regulatory decisions. We're using
MELCOR to come up with a source term for 50.44,
combustible gas control.
We're using RELAP and TRAC to modify 50.46
ECCS.
There was another subcommittee out of OSU,
and you heard that we're using RELAP and CFD and REMIX
to do PTS.
We're using RELAP SCDAP to do steam
generator high temperature severe accident work, and
you had a separate briefing on that.
We did some work on combined injected rod
LOCA as part of the CRDM issues, and we used RELAP and
PARCS and for AP1000 we'll be using RELAP and TRAC.
For our work on synergy we're going to be using TRAC
and PARCS.
So we're actually using these codes to
make regulatory decisions, and that's very healthy.
And much of that work is being done in-house, and
that's very healthy.
The last point that I wanted to make is
that in prior years it was typical to have a vendor
come in with some calculations and what we would cause
our contractors to do some calculations to check
vendor calculations. But the regulatory decision, to
a great extent, was based on what the vendor came in
with. But for changing the rules, we're looking at
this synergy issue or the objective art issue. These
are safety issues that are before us. And we're using
our analysis to make those regulatory decisions. We
don't have a vendor to balance this stuff off against,
except in the AP1000 case. And that puts a greater
burden on us.
The entire Office of Research is paying
far greater attention to QA than it did in the past,
because we're using this for regulatory decisions and
we're trying to do the code development and
configuration management, etcetera, to modern
standards in support of those regulatory decisions.
It's the first time that somebody gave me
a microphone. I didn't realize I talked softly.
With that short introduction, I'd like to
turn it over to Jennifer for about an hour -- Jennifer
Uhle who is the assistant branch chief now in our
branch.
MS. UHLE: We're going to do something a
little different. Usually we talk about the status of
our code development efforts and then talk a little
bit about applications. But I think you guys are
tired of hearing it in that order, so today I'm going
to start off with what we're currently using our tools
for. The question gets asked what do you use the
codes for; they're, of course, time-consuming to
develop and we have an invested effort in that. And
so we're going to be answering this question for you,
hopefully.
So I'm just going to talk a little bit
about the branch mission, the current uses of the
codes at NRC. You know, the current applications we
have for licensee submittals, generic issues, risk-
informing regulation, design certification. And we
just draw the conclusion that you'll find on the
summary side now, and that is that you do utilize the
codes, they are used at NRC for field application.
And it is our goal to continue to improve this
analytical capability to respond to these emerging
issues.
We always discuss about the consolidation
effort. That effort, of course, sometimes gets in
trouble for the fact that we are not making
improvements to the physics as quickly as some people
may want. And I just want to focus or make the
statement that we are consolidating first, we are
making improvements as we need to respond to these
applications as they arise. But by the end of 2002
we'll be in a prime position to have one code. At
that point in time we'll really focus on improving the
physical models and as well as the very detailed
developmental assessment. And Joe Kelly and Steve
Bajorek will be discussing that further.
I'm going to skip over this, because I'll
do that on the summary side, but Jack Rosenthal had
pointed out that we do have five recent hires that
have really added to the capability of the branch, and
you'll be hearing from them.
MEMBER LEITCH: Jennifer, maybe it's
obvious, but I'm not sure I understand. What is the
advantage of a consolidated code?
MS. UHLE: We used to have four thermal-
hydraulic codes. And we used RELAP for PWR small
break loss of coolant accidents and transients. We
used the TRAC-B code for large break loss of coolant
accidents for PWRs. We used the TRAC-B code for BWR
applications that only required one 1-D kinetics. And
then we used the Ramona code for places that required
3-D kinetics capabilities. And because of that each
of these codes have very similar features. They're
not that different, and so we had a lot of maintenance
points; that wasn't an efficient way to operate. It
was more costly than it needed to be. So when we
needed to make improvements we, in a sense, had to do
it four times over. So improvements weren't made as
fast.
Additionally, the user base was
distributed across these four codes. So, again,
instead of moving forward we were sort of moving in
parallel and not making improvements as fast as we
would have liked.
Additionally, each of the codes had a
different input deck. And so when you're looking at
maintaining these large input decks, these very
complex models, you would have to do it for two
different inputs, because the PWRs would use RELAP and
TRAC, the BWRs would use Ramona and TRAC-B. So it
just wasn't an efficient way to proceed, especially
with the budget reductions and the fact that we wanted
to bring the technology in-house and have in-house
staff to develop and maintain and use the codes for
the regulatory applications.
So that was the decision to go with the
consolidated code. And what we did is we selected
TRAC-B as the base of that consolidation, and we
modernized it so it's a new architecture. It's very
easy to modify, very easy to extend to other
applications and to couple to other tools like a CFD
code. We haven't done that yet, but this is where
we're heading.
And what we've done is we've just taken
the features that were in the different codes, all of
the four different codes, and we've only taken the
different things that the other codes could do and
brought them into TRAC-B. We now call it TRAC-M
because it's modernized, and we're trying to find a
name for the code. It's a very sore subject.
MEMBER LEITCH: And this will be
consolidating code that NRC --
MS. UHLE: Yes. Yes. It's in-house
expertise. We work with contractors. Gil Actess is
in the back of the room. He's at ISL, Information
System Laboratories. John Mahafty is Penn State
University. He was an original developer of TRAC-B at
Los Alamos. He's at Penn State now. He is our
numerics guru. And Tom Downar at Purdue University is
working on -- is more of the original developer of the
PARCS code. Now we don't use the code as stand-alone
in PARCS; we've coupled just the kinetics routines to
TRAC. So it's a modular, so it's the PARCS modular.
But, again, we work alongside of the contractors, the
staff does, and we've really developed in-house
expertise. Tony Ullses is now starting to do PARCS
development so that we can rely more on in-house staff
and rely on contractors for specialized skills, so
it's not part of the staffing plan to have a full-time
employee on one of those particular skills.
MEMBER LEITCH: Now, when I think of the
consolidated code, I think in terms of simplicity and
efficiency. That raises sort of the feeling that
maybe there's some compromise of precision for the
individual codes, a specific code for Bs and Ps and
small break. large break and so forth. Is any of that
precision compromised?
MS. UHLE: That was a concern. I think
Dr. Shack is of that mind. I think that Dr. Zuberg
was of that mind as well. And the bottom line is we
just couldn't continue to operate that way. We
couldn't make any more improvements to the codes
because we were spending all of our resources on
maintenance. And so as these issues were identified,
we just didn't have the staff or the budget to be able
to make the changes. So in a perfect world maybe that
would be the best way to go, if you had infinite
resources and infinite time.
So the consolidation plan is that we are
forming the consolidation activities. We can read all
of the input decks from all the other codes, so we've
recovered the input decks. And at this point or
shortly we'll be starting the developmental assessment
phase. And for the targeted applications of each of
the predecessor codes, we will be comparing the
results of the modernized code to the other codes to
make sure that we're as good as the other codes for
those applications.
And the way that the architecture is set
up, it's really the physical models; wall drag,
interfacial drag, interfacial heat transfer, et
cetera. Those are the points that made the codes
different. And, for instance, the solution of setting
up the numerics, solving the matrix, performing input
processing, performing or exporting the answer to a
graphical tool; those are all common things. So
really the only big differences between the codes is
the physical models. The architecture of the
modernized code is allowing us to do component-
specific physical models.
If I'm a pipe, I'm going to use this
interfacial drag, this wall drag. If I'm a channel
component in a BWR; okay, now I have a rod bundle
there, the interfacial drag is going to be different
than it would be in a pipe of the same hydraulic
diameter. So because of the architecture it's set up
to very easily incorporate component specific physical
models which will allow us to be as good, and then
eventually better than the old codes. So we're
accommodating that concern.
We have to prove that to you, but that is
our goal.
CHAIRMAN WALLIS: Of course, you're also
checking that this pipe really is a pipe and isn't a
pump because --
MS. UHLE: Yes, of course.
John wanted to say something. John
Mahafty.
MR. MAHAFTY: This is John Mahafty.
I'd just like to make a comment. I've
been kicking around with computers since they took up
the whole room and they had the kind of computer power
you have in your watch right now, and I understand the
concerns about efficiency from that kind of ancient
perspective. But the fact is now memory on computers
is massive and it's cheap. Disk space on computers is
massive and cheap. So that it doesn't matter to me if
I've got a large code with a bunch of special
subroutines for interfacial guide and BWR and another
set for interfacial drag and PWR core; if I'm running
BWR, that stuff never gets swapped in where the action
is, which is your real local memory on your chip. It
sits out somewhere and gets swapped into virtual
memory. It's not impacting you from the standpoint of
the efficiency of operation of the code, but it's
there when you need it and it's tied together with all
these things that everybody needs to make the
maintenance and the improvement of the package
important. And things don't get overlooked as much.
Now I can remember the old days. It used
to drive me nuts. We'd find some problem with TRAC-E
and we knew that it was an important issue that the
people in BWR side, maybe we only had five of them
looking at it and communicating that and getting all
of this information to run off, it would sometimes
take years. But now it's in one place and there are
people thinking about it as a whole, so you don't lose
improvements that are applicable to everything, and
it's a big advantage.
MR. ROSENTHAL: Let me do a follow-up
then, if I may. And, John, you're absolutely right.
Every time we turn around, of course, it's a tenth of
what it did before for more horsepower, computer
horsepower. But people are expensive, and it takes
over a staff-year to create an input deck, one of
these really big input decks. And so you really gain
some efficiencies by being able to use decks that were
previously created or have common decks for purposes,
etc.
So I think we're really going to achieve
some efficiencies.
MS. UHLE: I just want to add to the idea
of taking a year to develop a plant deck, and that was
again in Jack's old time frame, in the olden times, in
the time of the dinosaur. That's how long it used to
take.
I'll be doing a bit of a presentation on
the graphical user interface, which we've also
recognized the inefficiency associated with plant
modeling and feel we have a program to handle that.
We've demonstrated that before to the ACRS, but I'll
be touching on some of those points that I think bring
that to light, that we have improved the efficiency of
the plant modeling.
MEMBER SHACK: Just out of curiosity.
What language have you settled on? I mean, these were
originally --
MS. UHLE: Fortran 90.
MEMBER SHACK: Fortran 90.
CHAIRMAN WALLIS: Well, the plant modeling
involves people looking at a lot of drawings and then
turning this into computerese. I would think with a
new plant and the plant is already a computer model
before it's even being built and you don't have that
problem; having to look at drawings and figure out
where the pipes go and --
MS. UHLE: You're assuming that we
communicate to the licensee.
CHAIRMAN WALLIS: Well, if that's the
problem, you need to fix it.
MS. UHLE: At this point that is something
that we have thought about, being able to scan in
something from the architectural engineers.
CHAIRMAN WALLIS: That's the way that
industry does it.
MS. UHLE: Right. Right. Well, they
don't build an input deck by scanning in the graphics
--
CHAIRMAN WALLIS: But in the automobile
industry, if they want to get a piece from a supplier,
they just send them a computer model of the stuff that
they need to know and they've got it.
MS. UHLE: But the computer model's not
going to have lost coefficients, reverse and forward
lost coefficients. I mean they're going to have
geometry, and that's what we can recover.
CHAIRMAN WALLIS: So you have to figure
that out.
MS. UHLE: But the rest of it is going to
still require somebody knowing the code, knowing what
each of the input is required.
Well, I mean we have talked about that as
an ideal way to go, being able to recover any of the
geometric information.
CHAIRMAN WALLIS: Right.
MS. UHLE: We have talked about that.
MEMBER SHACK: They probably don't have
computerized geometric models in most of these plants.
MS. UHLE: They'll have like tech -- what
is it called -- CAD drawings. They'll have CAD
drawings. And so we've thought about being able to
take in the data from the CAD drawings and getting the
geometry. And that is somewhere we want to head,
we're not there yet. And of course, at NRC we don't
have CAD drawings, so it would require interface with
the industry.
I want to talk to you about the mission of
the branch, to give you an idea that this is the
Safety Margins and Systems Analysis Branch. So we are
tasked with the idea of maintaining these analytical
tools. We're also tasked with maintaining the
infrastructure for the understanding of the
phenomenology to help out NRR on more complex issues.
And this is applied to severe accidents as well as the
thermal-hydraulics, and as well as the field behavior.
What you're hearing from us today is the thermal-
hydraulics, but we are hoping to follow suit in the
severe-accident and field-behavior areas so that the
team can seamlessly interact throughout the branch;
and that includes coupling the computer tools, the
field behavior code to the thermal-hydraulics code,
the severe accident code to the thermal-hydraulics
code, and bringing in-house expertise. And so it's an
exciting time in the branch.
Hopefully, if all the good things you hear
today, you can think that's going to be applying to
severe accident. And if it's something you don't
like, well then tell us so we don't make the same
mistake twice.
MEMBER KRESS: When you say criticality
safety, what all is wrapped up in that?
MS. UHLE: Criticality safety originally--
well, for instance in the dry cask PRA they have asked
the branch --
MEMBER KRESS: Okay. You're not just
limiting this to reactors then?
MS. UHLE: No. No. For instance, the
most --
MR. ROSENTHAL: The burn-up credit comes
from in our branch analytically and we provide, as a
user need --
MEMBER KRESS: I understand what you mean.
MS. UHLE: Tony Ullses, in the back of the
room, is currently running some calcs with the dry
cask PRA to just double check that there's, obviously,
very -- I don't know what the word is -- low, low, low
probability that anything could happen and cause a
criticality accident. And so he's doing that in the
branch, because we have the reactor physics tools.
We've coupled the reactor with some kinetics tools,
but we're getting the reactor physics and with that
there is quantum PYLAR codes for criticality.
MEMBER FORD: You mentioned safety margins
on this slide. Is there any plans in the future to
incorporate, for instance, aging phenomena for
construction materials?
MS. UHLE: We are going to talk -- I will
actually talk a little bit about that with respect to
the power uprate synergy program that we're undergoing
at this point. And Joe Staudenmeir is the lead on
that. But, additionally, we do interact with the
engineering division, well for instance, through the
pressurized thermal shock rule we are looking at risk
informing the PTS rule in the way that we're giving
them thermal-hydraulic information and then they're
putting it into their FAVA code for the fracture
mechanics.
So the whole office, really, I think it's
a nice tie and we're all starting to interact a bit
more. There's a lot of cross-division, cross-branch,
as well as in the branch cross-section interaction.
MEMBER FORD: Forgive me, because I'm new
to this organization. Is this a new mission or has
this been a mission you've had for ten years?
MS. UHLE: I think this is a mission that
we've always had, but I think the way NRC is currently
operating we're trying to do it in a more efficient,
more --
CHAIRMAN WALLIS: Integrated.
MS. UHLE: Yes, integrated and more of a
outcome-oriented, and all these management buzzwords
that make you sick. But, you know, looking at the
user offices as our customers, looking at the fact
that we're supporting the PRA work as our customers.
And because of that, I think this has helped as far as
people understanding who is doing what and who to go
to talk instead of not knowing and calling their
professor or, you know, and not knowing what NCR is
currently doing.
CHAIRMAN WALLIS: It also helps if your
customer is really listening and is in on that
decision making.
MS. UHLE: Oh, yes, right, and that goes
to the user need process.
Okay. So we're getting into the activity,
because now I'm talking about power uprates synergy.
So you read my mind here.
This was actually I think at one point
discussed by the ACRS, the full committee, looking at
the potential for synergy. Synergy coming from the
fact that we're operating with higher burnups, higher
power and plant aging. And we are currently looking
at license amendments for BWRs up to 20 percent power
uprate.
Also the Office of Research -- I don't
want to be giving you a full review of this program
because I'm not the lead on this program, but I just
want to talk about our branch's use of the codes to
support this program.
We've got an independent study we'll be
doing; the best most rigorous method we could do would
be a level 3 PRA before and after the results, but we
don't have the time or the staffing, or the funding to
do that. So we're trying to do this in an efficient
way, so due to the time and funding limitations we're
going to focus on components and the scenarios of high
risk significance, and using the knowledge that we
have in the field to point to the things that are most
sensitive to the changes. We're looking at the
results of NUREG-1150 as a guide. And we're going to
be looking at generic safety issues and reviewing them
to see that if there was something that is affected by
any of these changes within the operations.
MEMBER KRESS: Now let me see if I
understand that. You will do a level 3, but for just
selective sequences?
MS. UHLE: Yes.
MEMBER KRESS: And those sequences will be
the ones you feel are more important?
MS. UHLE: Yes.
MEMBER KRESS: And you'll pick out a
number of plants to do this with?
MR. ROSENTHAL: The level 3 would include
consequence analysis.
MS. UHLE: Right.
MEMBER KRESS: Yes, you'll forget about
LERF and go to the full consequence.
MS. UHLE: Well, we're looking at
consequence on the synergy program after listening to
the advice from Joe Staudenmeier.
CHAIRMAN WALLIS: So you're going to look
at casualties in the surrounding countryside and
things like that? I mean is that part of your
mission?
MS. UHLE: Yes, I mean it's going to
result in a source term and then --
MEMBER KRESS: Well, before we get carried
away, I think I'd like to lend the Subcommittee's
support to your doing that. Because LERF can only do
it when you are talking about power upgrades.
MS. UHLE: Well, I mean, the focus is
looking at the source term.
MEMBER KRESS: Yes, absolutely.
MS. UHLE: And we are going out to source
term.
MEMBER KRESS: And we really ought to do
the level 3 in this case rather than stop at LERF.
MS. UHLE: And if we have source term
going to, you know, the health effects, I mean that's
-- I don't see how that's a big step.
MEMBER KRESS: Will you use specific sites
for this or some sort of a --
MS. UHLE: Joe, do you want to stand up
and talk? Joe Staudenmeier is the lead on this. I
mean, maybe a lot of it could just be my
misunderstanding, but I mean if we're doing source
term, I don't see why we wouldn't do the final health
effect. I mean, that's just a matter of running the
Max code, which takes 5 seconds. But maybe I'm
offering work that the office isn't willing to do. I
don't know.
MEMBER KRESS: Stick by it, I hope you do.
Go ahead.
MR. STAUDENMEIER: Joe Staudenmeier.
Tentatively we had planned to do
consequence analysis. We don't really have all the
details of this whole study all worked out yet, but
tentatively we'll look at the consequence analysis
with the PRA people. We're going to provide guidance
based on NUREG-1150 study on what sort of sequences we
should be looking at and also engineering is providing
information on materials and things like that.
MEMBER KRESS: Okay.
MR. STAUDENMEIER: It's hopefully going
to be an integrated study that gives consequence
numbers, or at least what we think may be resulting
consequence numbers being more like a prioritization
analysis rather than a full level analysis.
MEMBER KRESS: Would you use the SPAR
models for this or --
MR. STAUDENMEIER: I don't know the
details of what GRA would be
MEMBER KRESS: All right.
MEMBER FORD: Could you -- and you ought
to be able to put the government timing and funding
limitations off. In light of, for instance, synergism
between higher power flux and plant aging from a
physics point of view, there's a lot of things which
are not understood in a quantitative sense. So there
is a lot to do beforehand. So far as timing and
funding limitations would you have these for is it 3
years? --
MS. UHLE: I think we have funding out for
another 3 years. Is that right, Joe? Three years?
MR. STAUDENMEIER: Yes, the program is
scheduled to go over three years. The total
contracting money for the first two years, I think, is
about $800,000, and from last year about $1500.
CHAIRMAN WALLIS: So probably three years
most of these BWRs will already have had power
operation approved?
MS. UHLE: That's coming out of NRR. What
we're doing is an independent analysis.
CHAIRMAN WALLIS: I know, but it's so
interesting. So your report will come out after the
fact and then --
MR. STAUDENMEIER: We are working on a
confirmatory report and we are not going to concern
ourselves with the licensing process. Unless we do
find something. If we do find something it will
affect licensing, obviously, we'll provide that
information.
MEMBER KRESS: Better late than never.
MEMBER SCHROCK: So the BWR presentations,
these upgrades claimed that the bundle power is not
increased, and the flux therefore is not increased.
So you have a situation in which the total power in a
system is increased by --
MS. UHLE: Right.
MEMBER SCHROCK: -- working the bundle so
they're both hanging over the mark. But it doesn't
come through clearly to me how you're dealing with the
increased total power, I mean in the context of source
term and things of this sort. You don't have a higher
local power density, and so the onset of failures is
not changed in the sense of local conditions, but the
amount of the core that's involved in the onset of
failures is increased.
MS. UHLE: Right.
MEMBER SCHROCK: How is that --
MS. UHLE: Affecting source term?
MEMBER SCHROCK: Yes.
MS. UHLE: Well, I mean with the higher
power the higher fission productivity and then of
course if you're getting --
MEMBER SCHROCK: Well, of course. But the
issue is how much of it gets out.
MS. UHLE: Right. Right.
MEMBER SCHROCK: And how does the failure
propagate?
MS. UHLE: Right. But if we're looking at
on the very unlikely situation where you'd have a core
melt, then you know it's going to be the average of
the core that's determining the source term, not just
the hot bundle.
MEMBER SCHROCK: Yes --
MEMBER KRESS: You would get more out
sooner.
MEMBER SCHROCK: Oh, I'm sure you'd get
more out, but my question is how it's being determined
in these new evaluations.
MEMBER KRESS: Well, it depends on how
they nodalize the core.
MEMBER SCHROCK: Because as I read the
stuff that we received, I was reading there's an
increase in flux, there's an increase in temperature,
there's an increase in this and that, which we heard
in the arguments in favor of the uprates it didn't
exist because we don't have an increase in bundle
power, we don't have an increase in center line
temperature of the fuel, we don't have this, we don't
have that. Whereas, the description that I read
sounds to me like it's contrary to the claims that
were made in the evidence supporting the approval of
these 20 percent uprates.
MS. UHLE: I think Joe wants to make a
statement here.
MEMBER SCHROCK:
MS. UHLE: He's behind you.
MR. STAUDENMEIER: The source of the core-
melt progression in source term release is something
we're going to be evaluating under this program. We
plan on planning some severe accident calculations.
I think we'll probably be talking in more detail about
this program, coming up with a presentation sometime,
I imagine being the first half of next year coming
down to explain what the parts of our program are and
schedule a presentation just describing this in more
detail. Right now Jennifer has a long way to go, and
this may not be a good time to discuss it any further.
CHAIRMAN WALLIS: She has a long way to go
in terms of the slides she's going to cover. You're
going to cover 19 or 20 or so?
MS. UHLE: Well, I don't know. I'm
trying.
Along those lines, though, I just want to
point out that we will be using the codes in the
branches, the severe accident analyses with melt core,
talking about the melt progression and then the
thermal-hydraulic codes. And so we'll be focusing on
the risk-significant events and the risk-significant
components providing input as success criteria,
operator action times, stating the case of ATWS, and
also different component failure modes.
If it`s a DET the division of engineering
to look at the effect of additional hydraulic loads on
the components, crunch the numbers and come out with
a new risk value.
So I'll skip the next one there, because
I think we've talked about that.
One thing I want to talk about, though, is
the fact that we're using this code and how can you be
assured that we are getting an okay answer for, say,
the BWR cases. The next stage in the consolidation is
very consistent with the fact that we need to do a
developmental assessment. And so what we're going to
do is that we are focusing on the BWR models first.
We'll be looking at them in the consolidation matrix,
the DA matrix, the BWR models. And that's, of course,
good timing with respect to the BWR synergy. So we
will be running a developmental assessment for BWRs
with the code, and we'll be using the TRAC code for
that.
We are currently involved in the Peach
Bottom Turbine Trap using the TRAC-M code in the PARCS
3-D kinetics module. And Tony Ullses -- he's in the
back of the room -- he is the lead on that.
Based on the results, and I have a few
results for you to show you, we found that we know
we're going to have to do some BWR specific physical
models. And what was put in was an interfacial drag
model was changed and -- I think it was the two phase
loss multiplier for -- I'm sorry, the two phase
multiplier for the wall drag that was important in the
BWR sense.
Once we replaced those models and reran
the Peach Bottom Turbine Trip for just the CHAN, you
know the BWR channel component in the core, we got
very good answers. We're still looking at it again to
focus on what models we need to change to improve
those answers. And I just want to --
CHAIRMAN WALLIS: If you take the RELAP
models and put them in TRAC, do you predict the same
answers as RELAP predicted?
MS. UHLE: If we were to do that, it would
take time to do it. We haven't done that. But in
general -- in general you would say yes.
CHAIRMAN WALLIS: You would expect --
MS. UHLE: If we run in the semi-implicit
numeric scheme.
CHAIRMAN WALLIS: It's a test that we
probably should run, isn't it, so that there isn't
something peculiar about TRAC which gives different
answers from RELAP with the same models?
MS. UHLE: Well, that's where the
developmental assessment work will --
CHAIRMAN WALLIS: You haven't done that
yet?
MS. UHLE: That's what the next stage is.
CHAIRMAN WALLIS: I mean, it's related in
a way to Graham's question; when you consolidate these
codes, the question will arise probably about whether
or not you're recapturing what the codes could do
before.
MS. UHLE: Right. Right. And so that's
why the next phase of consolidation is the most
important phase.
CHAIRMAN WALLIS: So you won't really find
out if there's a hitch to consolidation until you get
to that point?
MS. UHLE: You have no faith.
MEMBER SCHROCK: I've expressed concern
for years about using interfacial drag as a tuning
device in the codes. And what can you say about what
you're doing now that's any different than what's been
done before? In terms of the physics, isn't it
necessary to have a clearer explanation as to why you
needed a different model --
MS. UHLE: Yes, we do.
MEMBER SCHROCK: -- for one reactor as
compared to the other?
MS. UHLE: Yes. In particular the CHAN
component. The CHAN component is essentially a pipe.
And if you put in your hydraulic diameter --
MEMBER SCHROCK: That`s in the code, but
in the reactors they're rod bundles.
MS. UHLE: I know. Right. Exactly. So
currently in the code, in the TRAC-M code, if you're
going to have a CHAN component, it is a pipe with a
different hydraulic diameter. So your interfacial
drag is going to be much -- you know, calculated to be
very high. Because in reality you have this channel
there -- sorry. You have this rod bundle there and
you know with the same hydraulic diameter you have a
much lower interfacial drag. So in that particular
instance we have to put in an interfacial drag model
that reflects the fact that there is a rod bundle in
this pipe. That's physically based.
MEMBER SCHROCK: Well, I didn't, I guess,
fully understand the argument.
In both reactor systems you have rod
bundles. You do have pipes. And so now you're --
MS. UHLE: In the PWR we have a 3-D
hydraulic model, so it's not a pipe because the
hydraulic is three-dimensional -- you can have cross
flow, what have you, you don't have the channel boxes.
MEMBER SCHROCK: Well, there's a scheme
for accounting for cross flow. Calling it three-
dimensional is a stretch.
MS. UHLE: Not in the TRAC code. It's a
three-dimensional model, three-dimensional hydraulic
model.
MEMBER KRESS: They don't use these little
--
MS. UHLE: We don't use the cross flow
connections.
I think Joe Kelly wants to say a few
words, maybe clear it up.
MR. KELLY: This is Joe Kelly, from
research. And I think I can clear that up, Professor
Schrock.
In TRAC-P, its mission was for large break
LOCA. So consequently, interfacial drag in the core
for normal, you know, bubbly flow was never considered
a priority. They were always worried about reflux,
first of all, from boiling etcetera. So the models
that were developed for that actually were fairly
crude, based on bubbles and slugs where the slug size
is limited by the hydraulic diameter of the channel.
But, as you know, in an actual LOCA configuration the
vapor structures actually span a number of
subchannels, and it can lead to much higher slip than
you would get if you only took into account the
hydraulic diameter of a rod bundle.
So, because the modeling TRAC-P is
relatively crude, it was in fact never extensively
accessed against rod bundle void fracture data.
There's no expectation that it would do a good job.
And what we've found is, yes, indeed it does not do a
very good job when you apply it to BWR operating
conditions. And so we needed to implement a model,
and what we chose was the one from TRAC-B that
actually does try to model the interfacial drag in a
rod bundle. And that's what was done for Beach Bottom
Turbine Trip, and I'll talk a little more about that.
MEMBER SCHROCK: Okay.
CHAIRMAN WALLIS: When we review other
codes, we've been reviewing other codes over the past
few years, we get a stack of stuff like this, you
know, the documentation. All the equations are
spelled out, justified, and the verifications are
explained. Are we going to get that for your code?
MS. UHLE: Yes.
CHAIRMAN WALLIS: When do we get that?
MS. UHLE: End of 2002.
CHAIRMAN WALLIS: That's a long way.
MS. UHLE: Well, we won't know what
physical models we're putting in the code until the
end of 2002, when we've done the developmental
assessment to make sure.
CHAIRMAN WALLIS: Well, do you have a
draft --
MS. UHLE: We have a theory manual for
the--
CHAIRMAN WALLIS: If you had a draft
version of the theory manual or something, we might
give you some useful input before end of 2002. And if
we're going to raise any problems --
MS. UHLE: Are you offering?
CHAIRMAN WALLIS: -- the sooner we do it,
the better.
MS. UHLE: So you're offering to be a
contractor?
CHAIRMAN WALLIS: Well, it just turns out
that in a peculiar way we should never fault the ACRS.
We act as sometimes reviewers of these codes and we
find what look like -- not what I should call errors,
but --
MS. UHLE: Right. We have a theory manual
for the base TRAC-P code. We can provide that to you
as well as --
CHAIRMAN WALLIS: It might be useful if we
saw that before you think you've got the final
version.
MS. UHLE: Right. I'll report that back,
although my management is here now.
CHAIRMAN WALLIS: Because that would be
really embarrassing if we found an error in some
fundamental thing after you think it's final.
MEMBER SCHROCK: We used to complain about
the lack of recommendation on TRAC. And I remember at
a meeting in Saratoga Springs -- from Las Alamos I
guess. Said that the latest version was fully
documented and I said "Well, I've never seen it." And
so there was some correspondence between he and four
others.
I think that he's under the impression
that it's been reviewed by the ACRS. I don't think it
ever appeared at the ACRS.
MS. UHLE: Okay. Well, I mean that would
be very helpful to us if you're willing to do that.
CHAIRMAN WALLIS: But you're going to
write your own documentation for these facts, right?
you're not just going to pick some original TRAC
document --
MS. UHLE: We're going to redo what needs
to be redone, yes.
CHAIRMAN WALLIS: Right.
MS. UHLE: Sure. As our developmental
work has been proceeding, we have quality assurance
guidelines and we've generated more documentation than
you can imagine.
CHAIRMAN WALLIS: If you go back to the
very original TRAC documentation, such as it was, it
was extraordinary. It was extraordinary, and it was
a maze trying to figure out what was happening.
MS. UHLE: I think of "extraordinary" as
good.
CHAIRMAN WALLIS: Oh, no, no. It was
extraordinary. I'll try to choose a word that's
neutral.
MR. ROSENTHAL: Let me just chime in. I
think what's going to happen, the goal and reality
would be that by the time we're done, this code will
have more review and more scrutiny than anything else
out there with a large user community, both
domestically and internationally. And we share source
code as well as compiled code. And we put it to the
user community so that it will be far better reviewed
and understood than I think the commercial code.
CHAIRMAN WALLIS: I was just trying to --
and you might think about how the ACRS could be most
helpful to you in that process. We don't have the
time to read every line and all that, but as you know
we do look at selected parts of this code
documentation and assure ourselves that it's credible.
MS. UHLE: And I guess you're interested
in the momentum equation?
CHAIRMAN WALLIS: We want to be helpful.
The last thing we want to do is to shoot you down in
some way.
MS. UHLE: Yes. I think the --
CHAIRMAN WALLIS: And the last time you
want to do it is at the end of the process.
MS. UHLE: Yes, I mean if you're willing
to do that, it would be great. I would think that we
would be accepting that.
CHAIRMAN WALLIS: Why don't you think
about how we might be helpful there.
MS. UHLE: I'm not important enough to
make that decision. It's these other people.
MR. ROSENTHAL: Sure you are. Sure you
are.
MS. UHLE: You have the results here, I
think, in your slides. I'll just skip over them. If
you want to pursue them, because I think we're running
out of time.
MEMBER LEITCH: This Peach Bottom Turbine
Trip, is that the generator breaker openings or how is
this -- or does that make a difference? In other
words, we run them along in the turbine trips, is
that--
MS. UHLE: It was a task scheduled at the
Peach Bottom facility. Tony Ullses can elaborate on
that; he's the lead, as well as Bajorek helped out
originally. Go ahead.
MR. ULLSES: It was actually a cycled test
that they ran at the facility during coasting down
gradually from 100 percent power.
MEMBER LEITCH: Okay. From a 100 percent
power? You say they were coasting down? They
weren't--
MR. ULLSES: Actually they were at low
power and they -- they actually had multiple trips but
they were down in the 60 percent power when they
started the trip and they actually disabled the
initial stops on the valve position --
MEMBER LEITCH: Oh, I sure. Okay. So
they closed the stop valves at 60 percent power.
Okay. Thanks.
The other question I had related to that
was you mentioned that in the previous LOCA, and I
guess you didn't slide 5, but you referred to the
Brown's Ferry ATWS. I guess is that a full-blown
ATWS, or is that the partial ATWS that occurred at
Brown's Ferry in '76 or something?
MS. UHLE: This is just on the matter from
the BWR synergy.
MR. ULLSES: That was a partial ATWS.
MR. KELLY: Partial ATWS, yes.
MEMBER LEITCH: So you're not using a
full-blown ATWS for this reference here?
MR. KELLY: Well, we're doing the plant
calculations on a full-blown ATWS, but we're going to
start with -- the deck was developed for the partial
ATWS, which is what that was for, and there were some
modern calculations on a former ATWS that we were
evaluating and we're going to start off by rephrasing
those calculations on TRAC-M using that deck as a
surrogate high power BWR4 deck as a full ATWS.
MEMBER LEITCH: Well, I guess I'm just a
little confused as to why you would use the Brown's
Ferry rather than a full-blown ATWS.
MR. KELLY: Well, we are going to run a
full-blown ATWS. What Brown's Ferry had was a
development responsible partial ATWS but there`s
nothing in the input that would keep it running at a
full ATWS.
MEMBER LEITCH: Okay. Okay. Thank you.
MS. UHLE: Okay. I'm going to skip now to
the MOX fuel issue. I think we have talked about this
before, but this is the idea of developing our
kinetics capabilities to deal with MOX fuel.
The PARCS, the Purdue Advanced Reactor
Core Simulator, that's the PARCS. What we do is we
coupled to just the kinetics features in the code. So
we use it as a module. And we are improving the
kinetics module to be able to handle MOX fuel. We're
adding the ability to do any number of energy groups
because of the fact that plutonium has huge capture
and fission resonances, and the beta is much lower
than in uranium. So you have to be much closer
because -- you have to be much more accurate because
you can be closer to prompt critical.
The way that the MOX core will be run is
we will be, we think, be using eight groups for the
MOX assemblies and two groups for the uranium
assemblies.
CHAIRMAN WALLIS: These are delayed
neutron group of the N?
MS. UHLE: Yes, beta delayed neutron
fraction.
CHAIRMAN WALLIS: You only need two for U?
MS. UHLE: I'm sorry?
CHAIRMAN WALLIS: You only need two for U?
MS. UHLE: Oh, sorry. The groups. No,
these are two energy groups for --
CHAIRMAN WALLIS: These are energy groups?
This is something else you're talking about?
MS. UHLE: Well, N groups. Additional
energy groups. N groups.
CHAIRMAN WALLIS: I don't know what an N
group is.
MS. UHLE: That means N number of groups,
how many groups.
CHAIRMAN WALLIS: Well, group of what?
MS. UHLE: How many bins of energy the
neutrons can be in.
CHAIRMAN WALLIS: I see. I see. Okay.
Okay. Thank you.
MS. UHLE: Two fields of neutrons, like
the vapor and the liquid.
CHAIRMAN WALLIS: You have two groups
there and eight groups here.
MS. UHLE: Yes. I think Dr. Kress can
help you on that.
CHAIRMAN WALLIS: Well, it seemed funny,
but I mean I guess this is a subgroup -- sub-sub
title. This is a sub of the title. The neutron
fraction isn't a subtitle of energy groups. Okay.
Never mind.
There are new problems with MOX, so we
really can't be surprised.
MS. UHLE: Yes.
CHAIRMAN WALLIS: New neutronic problems.
MS. UHLE: I'm glad I got that across.
MEMBER SCHROCK: Your bullet on reactivity
difference due to mix of plutonium in the range is a
little confusing. Error in reactivity can be closer
to prompt critical in MOX.
MS. UHLE: Yes.
CHAIRMAN WALLIS: That's because of the
delayed neutron fraction.
MEMBER SCHROCK: You need a comma there
somewhere? Error in reactivity still can be closer.
MS. UHLE: Can be closer to prompt
critical.
CHAIRMAN WALLIS: Well, you worry about
error because you don't have this cushion from the
delayed neutron, isn't that the idea?
MS. UHLE: Delayed neutrons. So your
prompt critical with --
MEMBER SCHROCK: Well, I understand the
problem, what I'm trying to understand is what message
am I to get out of this statement.
MS. UHLE: Okay. Take a step back here.
Okay.
Additional energy groups, there is a need
to have additional energy groups, more than just two,
that we currently use for uranium cores. Okay?
Why do we need additional energy groups?
We need them because of the fact that plutonium has a
lot of resonancy, and so around the epithermal range
and at the 1 eV range, and around the -- in Pu-241 you
get capture and fission resonances at the 1 eV to KeV
range.
So you have these resonances, whereas in
uranium you don't. You pretty much can bend your
energy groups of your neutrons into fast neutrons and
thermal neutrons because there's none of these big
resonances on the way scattering down to the thermal.
Additionally, you worry about error in
reactivity. We could have used the two energy groups
and --
CHAIRMAN WALLIS: Would you bend your
betas? The beta is an average of a whole lot of
different betas, isn't it?
MS. UHLE: It's an average of the betas.
CHAIRMAN WALLIS: And do you have to worry
about individual betas with plutonium?
MS. UHLE: Yes. In the 3-D -- yes. In
the code you do. I took off the 235 beta and the --
CHAIRMAN WALLIS: Well, that beta's just
an average for you, isn't it?
MS. UHLE: It's a beta for that isotope.
CHAIRMAN WALLIS: There are different
groups. Right. So there are different groups in the
beta --
MS. UHLE: Yes.
CHAIRMAN WALLIS: -- itself it subdivides.
Okay. You didn't worry about that now, because you've
got such a lower beta?
MS. UHLE: Well, in the fission event
you're -- I guess I don't understand what you're
asking. Do you understand what he's asking, Tony?
CHAIRMAN WALLIS: There are separate
groups. I get confused about the groups.
MR. ULLSES: Yes, the code itself, Dr.
Wallis, it actually on a node-to-node basis will
maintain an individual amount of the actual related
neutron.
CHAIRMAN WALLIS: But it looks at the
simpler fractions of the separate groups?
MR. ULLSES: Right.
CHAIRMAN WALLIS: Okay.
MS. UHLE: Just to give you an idea that
we have to be very accurate, more accurate than we do
in uranium cores because of the fact that we are
closer to prompt critical because of the --
CHAIRMAN WALLIS: So it's not just the
average, it's also the group which is slowest -- which
is governing in a rapid transit, isn't it? So it's
not just the average you worry about?
MS. UHLE: Well, it's the most dominant,
the most dominant group.
CHAIRMAN WALLIS: But, I guess Tony's got
it all under control. Tony's got it all under
control, certainly.
MS. UHLE: I'm sorry?
CHAIRMAN WALLIS: I said Tony has it under
control; that's all I'd really like to know.
MS. UHLE: All right. Great. So does that
explain this slide any better?
MEMBER SCHROCK: Well, no. The language
is what I'm criticizing, as in that statement error in
reactivity can be closer to --
MS. UHLE: Okay. Okay.
MEMBER SCHROCK: There is a reactivity
evaluation problem which is rather complex. POR is
big, it behaves pretty much like several critical
assemblies loosely coupled and each one has different
average values, of the delayed neutron fractions owing
to the fact that it has different composition at that
point in time, different weighting both the effect of
plutonium versus uranium neutronic properties and the
neutron fraction specifically. And so you're rolling
an awful lot of important information into a
simplistic statement here.
I've raised questions about this in the
context of other codes in the last year and I haven't
heard crisp clear answers to those questions. I don't
know that you're doing the calculation better than
some of the industry codes where they make claims that
they're doing it right.
Somewhere I'd like to hear a clear
explanation of how one keeps track of the local
compositions and how that information is then
impacting the calculation of such things as the 3D
kinetics. I haven't heard any of it yet.
CHAIRMAN WALLIS: You need to see the POX
-- you need to see the POX documentation.
MS. UHLE: We can provide that to you. We
have it written up, if you'd like that.
MEMBER SCHROCK: I'd like to see it.
MR. KELLY: Yes, we can do that. Sure.
MS. UHLE: Oh, sure. Or we could have a
separate briefing on the MOX development if that's --
MEMBER SCHROCK: You see, in the
documentation you're offering here, our code is like
the government's code, and therefore it's okay. You
guys can't challenge that because you've developed it,
it's your documentation and we're doing the same kind
of inadequate documentation as you do, but you've
judged it's good enough and therefore you've got to
accept the fact that we say it's good. What I'm
telling you is that it is not good engineering
practice. And I'm going to keep asking the question
until I hear some better engineering answers.
MS. UHLE: With respect to the MOX
capabilities?
MEMBER SCHROCK: It has to do with the
calculation of the reactor kinetics in a 3-D situation
in which the composition of the core is nonuniform and
evolving, it's different at different points in time--
MR. ULLSES: I understand. Right. Okay.
I can get back to you on that.
MS. UHLE: I mean I have a --
MR. ULLSES: I could take a stab at it now
or we can do it later.
MEMBER SCHROCK: No, I think we need to
get back.
MR. ULLSES: Okay. We'll get back to you.
MEMBER SCHROCK: Right.
MR. ULLSES: I`ll bring you the
documentation.
MS. UHLE: Okay. I can skip over the
other slide. I was going to get more into 3-D kinetics
methodologies for MOX, but I think we're going to have
a more detailed description of that provided to you at
a different date, if that's all right.
MEMBER SCHROCK: See, the term MOX is
generally interpreted as being -- as situations in
which the fuel is designed to be mixed oxide.
Whereas, what you really have in all reactors is some
form of MOX. And my problem with the calculations
that I see done is that this level of complication
gets getting short-shrift in describing what the codes
actually do. With the physics it is relatively
straight forward to understand in principle, but
complicated to deal with in the calculations.
MS. UHLE: Right. And I can tell you that
the way we're going to be handling the MOX cores is
that the uranium assemblies, the U02 assemblies, they
will be homogenized so that each -- the node -- the
power --
MEMBER SCHROCK: Have you asked yourself
the question of why does this issue of error in
reactivity arise when you're talking about mixed-oxide
fuel and not for reactors that have initial uranium
fuel?
MS. UHLE: Okay. With the reactor
physics, I mean you get three different types of
errors -- well, I mean stemming from three different
phenomena.
One is the number of energy groups that
you have because, of course, there are -- you don't
want to get into this.
CHAIRMAN WALLIS: No, his question is
different. I'm sorry. He said why is MOX different
from regular reactor because when you've got high --
MEMBER SCHROCK: In principle it's all
MOX.
CHAIRMAN WALLIS: -- burnoff, there's a
lot of plutonium there already.
MS. UHLE: MOX is because you're going to
have uranium dioxide fuel assemblies sitting next to
a MOX of plutonium dioxide assembly.
CHAIRMAN WALLIS: So there's increased
heterogeneity?
MS. UHLE: And so -- and you get very
different energy spectrums coming out of the plutonium
side because of the different cross sections for the
resonances. And so you get this very strong neutron,
this gradient in neutron flux between the assemblies.
MEMBER SCHROCK: Clearly the more you
complicate the spacial variation in fuel composition,
the harder the calculation becomes.
MS. UHLE: Yes.
MEMBER SCHROCK: And in mixed-oxide fuel
meaning that you have bundles of different composition
loaded into the reactor initially, it's going to be
more complex then if you load it uniformly and let it
generate its nonuniformity as it burns up. But you
get the same phenomena occurring to different degrees.
The relative consequences become more important when
you're talking about what you're characterizing as
mixed-oxide fuel cores.
MS. UHLE: The orders of magnitude --
MEMBER SCHROCK: But the phenomena are
always there.
MS. UHLE: Right.
MEMBER SCHROCK: And the codes need to
deal with the phenomena.
MS. UHLE: They deal with the phenomena.
MEMBER SCHROCK: Yes. My question is how
do they deal with the phenomena.
CHAIRMAN WALLIS: I think that's where you
have to look at the documentation.
MR. ULLSES: Yes, I understand the
question, Dr. Schrock. I mean, I can go through an
excruciatingly long discussion right now about
hydrogen --
CHAIRMAN WALLIS: I don't think we need
that. I think --
MEMBER SCHROCK: What I'd like is to be
given something to read that tells the story in a
clean cut fashion.
CHAIRMAN WALLIS: So would you agree to
give him something to read and then we can move on?
MS. UHLE: Yes. That is an action item
for us. By Monday we will have a clear --
MEMBER SCHROCK: Okay.
MS. UHLE: We have it written up. It's
upstairs. It's upstairs. We can go get it if you want
it.
CHAIRMAN WALLIS: Okay. Let's move on.
MS. UHLE: We`ll give you a brief
tomorrow.
CHAIRMAN WALLIS: Let's move on.
MS. UHLE: Why don't we go get it.
MR. ROSENTHAL: Why don't we provide him
with the documentation, okay. And then after he's had
an opportunity to look at the documentation, at his
discretion we'll schedule a morning session and we'll
talk about MOX. When we talk about MOX, we not only
talk about the physics, but we'll also talk the
neutron physics --
MEMBER SCHROCK: See, my emphasis --
MR. ROSENTHAL: -- we'll also talk about
source term and other related issues.
MEMBER SCHROCK: Jack, my emphasis is not
on MOX. It's on the fact that I look at old
documentation, which continues to be referenced, and
what I find is that people say you do these things
with delayed neutron yields and there's a table of
delayed neutron yields for U-235 presented in the
documentation in the early versions of RELAP5, for
example. And nothing's said one way or the other
about does this deal with the problem that the core
contains some other fissile nuclides and what are the
delayed neutron fractions from those.
It's the latter that I'm concerned with.
Why did they get lost in the shuffle?
When I raised it in connection with review
of another code, I'm told that it's all done
correctly, you just don't view it in -- yes, right.
Well, I'll believe it when I see it in a clean cut --
CHAIRMAN WALLIS: So you're going to see
it, Virgil.
MEMBER SCHROCK: Thanks.
CHAIRMAN WALLIS: And we're going to move
on. You're going to satisfy him with some
documentation, otherwise the question will just come
up.
So, can we move on?
MS. UHLE: I think everyone was aware of
the control rod drive mechanism issue. The Oconee
Unit 3 spring 2001 outage, there were circumferential
cracking on the CRDMs. We looked at the idea that
there's this potential for a rod ejection because of
the circumferential cracking.
The question was raised that you could
result in, perhaps, an ATLAS because of the fact that
you have collateral damage with the CRDM ripping off
and taking out a bunch of the other CRDMs in the area.
So Research performed a worst case scenario
calculations on the off chance that for some very
improbable reason there was a full ATLAS. And we did
a 3-D kinetic, 3-D hydraulics model using the TRAC
code. Jack had said it was a RELAP, but we had used
TRAC with this because we, again, want to keep
exercising the TRAC code. And we used the Boron
tracking to determine the effect of the RWST injection
shutting down the reactor.
The results of this actually confirmed NRR
from the analysis that NRR had done with RELAP5. And
what it showed was that there was no new phenomena
identified bounded by the current design basis and no
fuel heat up was expected, no core damage was
expected.
We did this as part of a confirmatory
analysis for which that was an activity that we did.
One thing to point out was that based on
the results of in running these codes is that they,
again, there are still bugs in the codes. And one
that we found was with respect to the Boron reactivity
coefficient.
In the PWR people don't picture -- well,
typically you're thinking of normal operation, you're
not picturing any boiling in the core. And the
reactivity coefficient for the boron, assuming no
voiding and it was based on parts per million versus
parts of boron per parts of liquid. And so it could
deal with boiling. And what we have done is, of
course, change it to what it should be, which is moles
of boron per the volume of the cell that you're
talking about. And this was actually identified also
in the TRAC-B code as well for the point kinetics
model.
So, every time we use these codes it helps
us.
MEMBER LEITCH: So in your calculations
you assumed that there was a partial --
MS. UHLE: A full ATWS.
MEMBER LEITCH: Oh, a full ATWS?
MS. UHLE: Yes. And so you're getting the
heat up, you're turning back around and with the
depressurization you're injecting the RWST water with
the high boron concentration and it's shutting it
down.
MEMBER LEITCH: Okay. So even with the
full ATWS you're still reaching these same
conclusions.
MS. UHLE: Yes.
CHAIRMAN WALLIS: It's a full ATWS and a
LOCA at the same time.
MS. UHLE: Yes. And Tony also said the
network.
Steam-generator tube integrity. You've
heard that, a briefing on that before. I think I'm
going to skip that for reasons of time. You will
please note that we will be using the thermal-
hydraulic code in the branch to look at those DPO
issues.
Let me get into risk-informing activities
that we have in the branch and the use of the codes in
those areas.
Of course, I think that you understand
what we mean by risk-informing regulation. The
current activities we have with respect to thermal-
hydraulics is risk-informing the ECCS rule and the
pressurized thermal shock rule. So 50.46 for the ECCS
and 50.61 for the PTS.
You have seen or the full committee has
seen a briefing in our risk-informing of 50.46.
I wouldn't say that it's really risk-
informing, the activities are more looking at any
modifications that can be made to Appendix K based on
the industry's desire to reduce regulatory burden.
And Ron Lauben and Steve Bajorek are the technical
leads on this in the branch.
So what has been looked at as an idea to
look at the Appendix K evaluation models and note the
real conservatisms in the code, and based on better
science can we replace the oxidation model for heat
generation to Cathcart-Pawel, because Cathcart-Pawel
does a better job as far as the heat generation.
We also have better science now with the
decay heat curve of 1994 standard versus the '71
standard. We were looking at that as an option. We've
been running code calculations to get an idea of the
change in the PCT based on these changes going to the
'94 standard or using Cathcart-Pawel versus Baker-
Just.
MEMBER SCHROCK: I guess we're going to
hear more about that?
MS. UHLE: Yes, in detail.
MEMBER SCHROCK: In details, but in my
mind it's just kind of strange that suddenly there's
a large activity going on to revise what has to go
through Congress to get approval, I think. Appendix
K is in 10 CFR, it's got to be -- it's part of the
legislation is involved here.
MS. UHLE: Yes.
MEMBER SCHROCK: There are lots of
complexity, but the background that's covered,
evidently, in SECY 01-133 seems to be totally lacking.
I don't understand how a decision can be made that we
must deal with a modification in Appendix K without
the technical evaluation that leads to the decision to
do that. Where is it?
MS. UHLE: That's our stance, though, the
division position, Research position, and we've had a
discussion with NRR in this manner that we're leaning
towards the idea of not modifying Appendix K because
of the fact that we have found nonconservatisms in
Appendix K. And the person who came up with the 71
times 1.2 was very good because they accounted for
those, in a sense, conservatisms.
MEMBER SCHROCK: Well, I read that so I
know what it is.
MS. UHLE: So you're the one.
MEMBER SCHROCK: Well, I'm not "the" one,
I was involved.
MR. KELLY: One of the ones.
MS. UHLE: One of the ones.
MEMBER SCHROCK: But what I'm hearing and
what I'm reading isn't a very accurate account of
that; not that that's a terribly important thing. But
what I'm getting at here is why is a lot of activity
going on here to revise?
MS. UHLE: What is the initiative?
MEMBER SCHROCK: What is the impetus to
revise Appendix K?
MS. UHLE: Appendix K --
MEMBER SCHROCK: What is the technical
basis for it?
MS. UHLE: Well, why this started was a
petition submitted by NEI looking at replacing the
'71 standard with the '94 standard. And so the idea
of reducing unnecessary regulatory burden or --
MEMBER SCHROCK: They're totally different
things. You're comparing apples and oranges.
MS. UHLE: I think -- can I finish what I
was saying?
MEMBER SCHROCK: Yes.
MS. UHLE: I think it'll -- okay.
That's why this, we started looking at
this one here with this idea to a risk-informed Part
50 is where a lot of -- we were looking at changing
Part 50, changing the regulations under this risk
initiative, this risk-informing initiative. And this
work here was put in with that based on the petition.
MR. BAJOREK: Jennifer, can I jump in?
MS. UHLE: Yes, sure, Steve.
MR. BAJOREK: This is Steve Bajorek.
One of the things that we're trying to
deal with is accuracy in the various models; the decay
heat or Cathcart model versus Baker-Just versus the
expectation that those can be changed in an evaluation
model.
I think there's been a recognition in the
SECY paper that the '79 or the '94 standard is
technically better than the '71 decay heat standard,
more accurate with regards to more recent data. And
likewise, with the Cathcart-Pawel versus Baker-Just.
The expectation that seems to have been
raised in the SECY paper is that we can just simply
replace those in Appendix K. The work that we have
been doing in our branch has been twofold: (1) To take
a look at what do you need to go from this decay heat
standard to the '94, and there's more complications
involved in dealing with the uncertainties. Norm
Lauben has been looking at that. But the other issue
is to what extent do the present day Appendix K
evaluation models depend upon the conservatism that
was inherent in the '71 plus 20 percent to cover other
issues.
Now, when we start to delve into this what
we have been finding are things like downcomer boiling
and fuel relocation would result in increases in the
peak cladding temperature that would almost offset any
kind of benefit that would be gained with the 1971
model.
MEMBER SCHROCK: Well, do you really
believe that the people that drafted 10CFR back in the
early '70s brought the uncertainty in decay power as
taking care of unrelated uncertainties?
MR. BAJOREK: No.
MEMBER SCHROCK: No. Okay. So why is
that brought up as an issue here?
MR. LAUBEN: Norm Lauben.
There was an evolution and it didn't start
out that nobody thought the decay heat multiplier, as
you say, we dropped another degree but as time went on
different things were discovered that was discovered
that there was a larger conservatism in the '71 than
was originally thought, but at the same time there
were -- how do I want to say this -- there was
creeping reduction in conservatism in Appendix K
evaluation models that ate away at some of the
increased margin that was perceived as time went by.
So, people then began to think, "Ah, well
there is extra conservatism in the decay heat model."
But it truth at the beginning we did not believe that.
MEMBER SCHROCK: Well, yes, I think that's
a historical fact that people have thought that way
that expressed their view, etcetera.
MR. LAUBEN: Yes, right.
MEMBER SCHROCK: But it's not something
that's documented as a basis for licensing evaluation.
MR. LAUBEN: And in fact --
MEMBER SCHROCK: So it's not something
that has anything to do with issues of whether you're
going to change it or not.
Those rules were created when there was a
lot of information that was still, basically, unknown.
MR. LAUBEN: Right. Right.
MEMBER SCHROCK: And did a remarkably good
job under the circumstances.
MR. LAUBEN: And may be lucky, too.
MEMBER SCHROCK: I think it was --
CHAIRMAN WALLIS: Well, I guess, one of
the things said here is that it could change the
regulations and became more realistic about decay
heat; it would look good and industry would think they
had gained something. It turns out you've got to be
realistic about some other things, which take away the
gains from the decay heat.
MEMBER SCHROCK: Yes.
CHAIRMAN WALLIS: And so that it's not
clear that there's a gain to anybody by changing the
regulations, except the new regulations would be more
based on more realistic physics, and that's probably
a good thing.
MEMBER SCHROCK: Well, I've probably
gotten into this at the wrong time in our discussions.
I know you have a presentation coming up on it. But
it does seem to me the starting point is thrown at
this committee in a very strange way. This SECY paper
has not been reviewed yet by this group. Okay. I don't
know what in the world it says or why they think
there's a sound basis. All I hear is rumors to the
effect that it is something that was initiated by NEI.
MR. LAUBEN: Is that true with the
activities with the --
MEMBER KRESS: Yes, we have. Not this
subcommittee.
MR. BOEHNERT: Yes, not this Subcommittee.
The full committee of ACRS I think, because it's
handled under subcommittee.
MR. LAUBEN: Have you reviewed all the
other copies of this?
MEMBER KRESS: Yes.
MR. BOEHNERT: It was handled by another
subcommittee, that's my --
MEMBER SCHROCK: Well, what I'm
challenging here is why does the Research branch of
NRR get deeply engrossed in a lot of considerations,
it's obviously an expensive thing to do, to address a
problem which somebody has told them is a change that
has to be made? On what basis can a decision such as
that be made without the technical work preceding the
decision?
MR. LAUBEN: Of course the technical work
has to be done.
MEMBER SCHROCK: Yes.
MR. LAUBEN: And I think 01-133 says the
technical work must be done. And if the decision
comes that we shouldn't change it, then we won't
change it.
MEMBER SCHROCK: My concern was, pure and
simple, that this is something that's going to get
railroaded through despite everything. And you're
saying that it isn't true. All right.
MR. AYER: Well, let me jump in. This is
Charles Ayer from Research. Let me just to correct the
record a little bit.
The SECY paper we're not risk-informing
50.46.
MEMBER SCHROCK: Yes.
MR. AYER: It was looking at several
issues, part of which was the Appendix K model for
decay heat. The petition to change the decay heat
came along later, and that was just something that's
come in very recently, but that was not the driving
force --
MEMBER SCHROCK: Okay.
MR. AYER: -- the NEI submitted a petition
and the agency jumped up and ran off to limit it to
50.46. It came in subsequent wanted a simple change
on 50.46. This other effort to risk-inform, which is
also looking at the large break LOCA and loss of
power. But that effort had been going on and is going
on. At the onset we're looking at the technical basis
for the smaller needs that Jack's branch is working on
to see if you can incorporate '94 decay heat, to see
what other things would have to be incorporated and
perhaps be more realistic in the other areas.
But I just wanted to make it clear this
wasn't initiated because of a petition from NEI.
MR. LAUBEN: As a matter of fact, Paul,
you were at several workshops last year in which this
group was starting to deal with it, so you know, so
you knew this was augmented to the initiative.
MR. BOEHNERT: Yes, that's correct.
CHAIRMAN WALLIS: Well, I think the
message for us is I think we would have said that you
could sort of change this decay heat code independent
of all the other considerations. And let's do it,
it's an obvious thing to do under the ACRS initiative.
We ought to follow that line. And what we are being
warned about here is if you do that, you're giving up
some conservatism which you really need to cover some
of these other things, and therefore you should be
more careful about saying, viewing the decay heat code
as something completely independent that you can fix
and then you can deal the other part separately.
MEMBER SCHROCK: Well, another way of
looking at it is that there is a very simplistic
rather conservative scheme for licensing put in place
in the early '70s that's antiquated, it was
grandfathered when the new rule was passed in '88.
And now the issue is, does it make sense to reduce
conservatism in an antiquated method. That's an
overall issue, it seems to me, and it needs to be
addressed, and it ought to be addressed by this
Committee, too.
To me it makes no sense whatsoever to say
we are going to go back and take all the conservatisms
out of an antiquated scheme and expect that it's going
to be technically sound in the end.
MR. ROSENTHAL: We briefed about two weeks
ago.
MR. KELLY: Two weeks ago, yes.
MR. ROSENTHAL: We briefed the PRAs and
members of the subcommittee --
MR. LAUBEN: And this subcommittee, too.
MR. ROSENTHAL: And this subcommittee.
MR. LAUBEN: We briefed three
subcommittees.
MR. ROSENTHAL: And now we have some more
technical work to do, and we would more than welcome
an opportunity bringing the technical work before this
Subcommittee. I think it would be very appropriate.
MEMBER KRESS: But I think from the point
I've heard in these other reviews that we're basically
on the same page you are with respect to that issue.
They're not going to just go in and blindly change
that Appendix K. They're going to look at what the
implication are.
And so I think we're closer to your side
of the table than you might think of.
CHAIRMAN WALLIS: We probably have to move
on.
MEMBER KRESS: Yes.
CHAIRMAN WALLIS: And we're going to have
a whole meeting on 50.46 some day.
MEMBER KRESS: Right.
CHAIRMAN WALLIS: And we can't dig into
that in depth today.
MEMBER KRESS: Yes. Okay.
CHAIRMAN WALLIS: But we've been warned,
I think, that we've got to worry about some of these
things, which has been very useful.
MS. UHLE: The point of slide too is to
point out that with respect to your concern about
doing the technical work to make sure that this is a
viable technical approach is that we will be running
and analyzing a great deal of cases with respect to
any of these activities. And it was with support that
we have given to NRR concerning the effect of the
downcomer boiling, especially as being a primary
concern that is shaping the technical position that is
leading in a direction that I think is very consistent
with yours. So we are using these tools for their
purposes.
Again, in the future we would also be
using them in the SECY paper to look at certainly the
effect of redefining large break LOCA size, looking at
success criteria evaluation for the PRA runs and the
effect of the different restrictions concerning delay
diesel generator start time, loss of offsite power and
signal failure. But, again, we will -- all of these
activities 54 -- or the risk-informing Part 50 are
going to be made, you know, using the available tools
and, as well as the knowledge and the analyses of the
staff.
MEMBER KRESS: My next door neighbor in
Oak Ridge has asked me to be sure you pronounce his
name correctly. It's Pawel, Dr. Pawel; just as if it
were P-A-U-L.
MS. UHLE: Pawel.
MEMBER KRESS: Yes. Not P-A-W-E. It is
spelled correctly, but it's not pronounced Powell,
it's Pawel.
MS. UHLE: I know how he feels, because
nobody pronounces my last name right either.
MEMBER KRESS: I know this is trivial, but
it upsets him.
CHAIRMAN WALLIS: Jennifer, are you going
to take a long time now?
MS. UHLE: No. I can skip over 61. We're
doing the same thing with 50.61. We're running the
tools; you had a briefing on that. We've made sure
that the calculations are consistent with data that
was taken at OSU. So we're looking at the idea of
when we use these potent codes how can we prove that
-- or at least appease the masses that the answers
that we are generating are acceptable. We're not
believing everything that comes out of the code, that
we're skeptical about it.
With respect to AP1000 design
certification, we had an NRR user need request
concerning looking at the Westinghouse assertion
concerning the scaling of AP1000 is consistent with
the AP600 work, and that they're claiming no
additional testing is required and minimal code
modifications would be required. That's the
Westinghouse position.
So NRR requested technical assistance from
Research to review these assertions, identify what
code versions should be used if phase 3 were to take
place.
And we for the small break loss of coolant
accident, I know a lot of you are involved in the
adequacy assessment of RELAP5 over that 5 to 6 year
period. TRAC had not been -- we didn't have a program
to do adequacy assessment for small break LOCA on the
TRAC code, so the RELAP code will be used for the
AP1000 phase 3 for small break. And in phase 3 if it
were to come in, the TRAC code would be used for the
large break LOCA application.
Now, one thing of note is, and an activity
that has stemmed from this initiative is that the
AP600 had a lower power density. So PCT values
predicted by TRAC were below the limits, the 2200.
AP1000 has an increased power density. We
realized that there won't be as much margin there and
we're based on calculation run with the reflood
models in TRAC. We're expecting that it would be over
the limit, not because of the actual physical
processes but because we have a lot of conservatism in
the TRAC large break model.
To remove some of this conservatism, will
we do a preliminary or an interim model development on
the reflood model. Bajorek is working on that
currently with Weidong Wang of the staff. And it is
hoped that or it is the goal to have that in by, say,
the spring or the summer and start doing some
developmental assessment work on that version for the
consolidated code. So by the time the consolidated
code is finished we will have, you know, this interim
reflood model developmentally assessed and use that
for the AP1000 submittal. Because RELAP large break
model tends to be nonconservative and TRAC is too
conservative.
Now, I don't want to confuse you with the
fact that the RBH, the rod bundle heat transfer
program. That's focused on developing a mechanistic
model for reflood. And we're thinking 2004, 2005 time
frame for it to be the model in the code.
What we're doing for the AP1000 work is
more of a -- we're simplifying what's currently in the
code with something that's more of a -- Joe, do you
want to say what you're doing?
I don't want to say it's simple, but it's
not the mechanistic model with a droplet diameter and
the interfacial area tracking, and what have you.
It's going to be easier to follow than what's
currently in the code. It will get rid of the
conservatisms that are coming from too much
entrainment at the punch front. And we're hoping to
have that done by the spring/summer time frame.
Do you want to --
MR. KELLY: I'll have several slides in my
presentation, so I'll wait for that.
MS. UHLE: Okay. I didn't mean to say
what you're doing is simple.
Again, we know that the phase separation
model in the RELAP5 code was determined to be
inadequate for the phenomena. It turned out that the
AP600 had so much water reserve that it didn't make a
difference in assessing collapsed liquid level, so the
code was determined to be adequate for the AP600
calculations with the fact that they have a higher
power density and the inventory to power ratio of the
AP1000 is reduced. We realize that the phase
separation model for the stratified conditions during
ADS 4 time frame is going to be of higher priority, so
we're looking at that.
Steve Bajorek will talk about that in more
detail.
We have there, too, for the PBMR design
certification that we're expecting to come in. I
think you know the background on that with the idea
that it is now a helium cooled/graphite moderated
reactor. It's a little bit different than the light
water designs that we currently deal with. It's a
pebble bed rather than the force flow parallel to the
bundle situation that we currently deal with.
We've drawn the conclusion that we would
be upgrading -- or not upgrading, but extending the
TRAC code and the MELCOR code to be used in real
certification if it were to come in. And we have
identified what needs to be changed in the code, and
you have a list of them on your slides. I don't need
to go into them. I don't think --
CHAIRMAN WALLIS: You're worried about
water ingress?
MS. UHLE: Yes, water ingress because of
the reaction with the graphite. Because you have the
second -- well, you have the cooling on the -- you
have the bring cycle but you've got the compressor in
the intercooler.
CHAIRMAN WALLIS: The intercooler is a
water cooler?
MS. UHLE: Yes, so you can get water
ingress. We can do water ingress and air ingress at
the same time and we have it working for the helium.
So we'll be able to run the whole gambit of the
accident scenarios with respect to the pebble bed.
Now, we do have some code development to
do as well as benchmarking, and we'll be doing that
in-house as well at Las Alamos National Laboratory.
CHAIRMAN WALLIS: Hydrogen and CO --
MS. UHLE: Yes.
CHAIRMAN WALLIS: -- process, or whatever.
MS. UHLE: Yes. Modifications also had to
be made to MELCOR and we've identified those, and
those will be done at CND and National Laboratory with
staff involvement.
So that's where we're heading. We're
going to not get away from this. It was thought that
maybe we would use a special code for the pebble bed.
Again, we're focused on this idea of having modules
that only need to be exercised if they need to be
exercised to get this approach with the consolidated
code having one code.
MEMBER KRESS: Well, what's the purpose of
looking at the water ingress for example. There are
no graphite structural ingress in there, are there?
MS. UHLE: There are no what?
MEMBER KRESS: Structural ingress in the
graphite? There's only the spheres of graphite isn't
there, they're not structural. So that's the --
MS. UHLE: It's a fuel damage issue.
MEMBER KRESS: We're looking to see
whether in the break the spheres --
MS. UHLE: Yes, that would be in the
severe accident situation.
MEMBER KRESS: -- break or something of
that kind?
MS. UHLE: Or would oxidize, getting
brittle, break and then get the fission products out
because the pebbles are the --
MEMBER KRESS: But you have no data on
sphere strength. I don't understand what you will do
--
MS. UHLE: That's the last bullet. Data
for benchmarking.
Originally in the budget this year there
was going to be some money for fuels testing.
MEMBER KRESS: I don't understand --
MS. UHLE: As the submittal comes in, that
will be ramped up to meet the data needs. We're not
going to use the code unless it's assessed.
MEMBER KRESS: You're going to degrade
these spheres, make them go through the separating
devise and see if the break --
MR. ROSENTHAL: All right. I'll be fast.
Presentations of the pebble bed say it's a very benign
system.
MEMBER KRESS: Yes. Absolutely. Okay.
MR. ROSENTHAL: And so we started asking
ourselves, okay, what about the accident provisions.
And bare in mind that design bases accident goes
beyond design base, or even that language is not yet
defined for this system.
MEMBER KRESS: Yes.
MR. ROSENTHAL: And we may be talking
about a spectrum of accidents, one accident, whatever.
Okay. And so we started saying, okay, what kind of
issues might we face, and we recognized that we needed
to start thinking about well what happens if we put
air in there, or water in there instead of helium, and
what kind of chemical reactions would take place, or
whatever. And because of the time it takes to develop
a code, we needed to get a jump start on these issues.
And that's really where we are now, you know, we
haven't thought it through. We're still defining the
research plan for it.
MEMBER KRESS: My question is --
MR. ROSENTHAL: But the concern is --
MEMBER KRESS: Yes. My question is are you
concerned about degradation of strength of these
spheres or are you worried about the effects on
fission products, or both?
MS. UHLE: Both. I mean, you get the
oxidation action causing fuel heat up and then you're
also getting fuel damage and how that's going to --
essentially if there's no containment, how the fission
products would be escaping because of that. So with
respect to the why in the TH code are we worrying
about that? Well, we're going to tell you how much
water comes in and what state it's in. Is it steam,
what temperature, what have you and get the oxidation
reaction and then, of course, going into the
theorizing for the core degradation.
CHAIRMAN WALLIS: So you're identifying
all the things that you're TRAC-M modification have to
be able to handle, that's really the message you're
giving us?
MS. UHLE: Yes.
MEMBER SCHROCK: What doesn't come through
clearly to me is why one would choose TRAC-M as a code
to analyze this new system.
MS. UHLE: It's probably at this point in
time --
MEMBER SCHROCK: I mean almost none of the
--
CHAIRMAN WALLIS: It's the only one they
have.
MEMBER SCHROCK: Well --
MS. UHLE: No, that's not right.
CHAIRMAN WALLIS: It's the only one they
will have.
MS. UHLE: No, that's not the answer.
I'll give you the answer.
MEMBER SCHROCK: But they're so different
from one another and --
MS. UHLE: Well --
MEMBER SCHROCK: -- all these gory details
of what goes on in water reactors has no impact.
MS. UHLE: Again, it's going to be
physical models that are going to be different. You
have -- I mean, if you look at the code as far as how
many hundreds of thousands of lines it may be, the
physical model package, I mean it's dinky. It's maybe
where the correlations are, maybe 400 lines or more,
or less. I mean, it's not -- putting in a different
wall drag or a different -- you know, effective
conductivity for the fuel. I mean, that's small.
What's in the code is the setting up of the matrix, it
is the communication of the data between the cells if
you have -- like in a 3-D. We have a 3-D vessel here
of porus media. The hydraulic model in TRAC is
essentially a porus media 3 dimensional model with
wall drag, that is assuming the flow is parallel.
Well, now the flow is going to be over spheres so we
have to replace that wall drag term with something
that represents the fact that you're flowing over a
pebble bed.
So, looking at all the codes that are out
there, TRAC was the one that had the less amount of
work done. We already have helium as a working fluid
in the code. Again, we have the porus media hydraulic
model.
We can do -- on the intercooler side, the
secondary -- if you want to call it the secondary
side, you know, we have the water loops for the heat
transfer. We have a turbine model that we have to
modify so that it's a two-phased turbine. But, you
know, we have the equation set up and already
dispertized; it's a matter of putting in different
physical models. But that's the, in some sense, the
easy part.
MEMBER KRESS: But you could have a break
in the intercooler, and the water is a lot lower
pressure than the helium. How do you deal with that
in terms of ingress to the water, or you haven't
gotten that far yet?
MS. UHLE: Well, I mean we are modeling
the -- you mean, the intercooler breaking and not
flowing into the helium, because there are two
different sections. And so if we had an intercooler
break, it'll just be like faster flow out and cooling
down; it'll be like a main steam line break in some
sense.
MEMBER KRESS: Yes, but I presume if
you're given a small leak, you know, have a crack in
it.
MS. UHLE: Yes.
MEMBER KRESS: And it gets some water
ingress.
MS. UHLE: Into the helium? Sorry?
MEMBER KRESS: How do you get water into
the helium is my point?
MS. UHLE: Oh, how do you get the water?
Well, for instance, if a steam generator were to
rupture, the same kind of situation where it's passing
over, if you get the water in -- oh, you're saying the
helium's higher pressure. Oh, I see.
MR. ROSENTHAL: Let's not get too far
ahead.
MEMBER KRESS: It's a technical issue.
MR. ROSENTHAL: At one time a few weeks
ago I asked at this plant if it had MSIVs, and I was
told, well MSIVs is the wrong term. There would be an
MHIVs. And so I said okay, is this plant going to
have MHIV? And I was told we don't know yet.
So let's not get too far out ahead of the
planning cycle. What we know is we started. We
really need tools to do analyses.
MEMBER KRESS: And that's the main thing.
MS. UHLE: But in the sense that you can
have a lower pressure or you can have a break in your
helium side, you get loss of forced circulation and
you still have hot graphite, you're at a low pressure,
water can get in. Because -- okay.
MEMBER KRESS: I'm sure there's some areas
that we can -- again --
CHAIRMAN WALLIS: We can't spend an hour
on the pebble bed reactor. We have to move on. Yes,
they're just giving us an overview, I think.
MS. UHLE: Yes. We will be using the code
and, again, the changes in the physical models are --
CHAIRMAN WALLIS: And you're thinking of
all the things you need to put in that code, you need
to build a model.
MS. UHLE: Yes, and we have done that or
in the process of doing that.
CHAIRMAN WALLIS: And when are going to be
ready to run?
MS. UHLE: Well, the work scope for next
year is putting in the physical models for next year
and finding data for benchmarking and doing modeling.
So by next time we meet in front of you, we should
have a pebble bed.
CHAIRMAN WALLIS: I just hope that you've
got models up and running before someone's already
made a decision for license on what the design bases
accidents are and all those sorts of things.
Do you actually have put in inputs to give
so quality decisions are made?
MEMBER KRESS: Well, is one of the models
going to be the fusion of water vapor in the graphite
spheres and what is the chemical reaction?
MS. UHLE: That would be the MELCOR side.
MR. ROSENTHAL: We may do that in MELCOR
fusion and hydrogen. We've got two major efforts.
One is TRAC and the other is MELCOR.
At one time we thought that -- just
conceptually that many of the pebble bed issues really
would be more chemical type issues and that the MELCOR
frame would be the place to focus. Then at the
experts meeting -- but we still had money in for TRAC.
Actually, it was Andy Kadak that kept bringing up
issues of reactivity events that might occur with
restacking or you lose the pressure, the walls move,
or stuff like that. Well, again, we had PARCS again
with TRAC. And so PARCS TRAC becomes the natural
place for us to want to explore that. But we really
are at the level of building the MELCOR models,
building the TRAC as tools for what we don't know yet.
MS. UHLE: I just want to point out with
looking at the kinetics, since that's been brought up,
we're really benefitting from the MOX program. I
can't believe I'm bringing that one back up. But the
things that are immediate is the soon to be needed for
the pebble bed work is similar to what has already
been for MOX. And so what has to be done for MOX is
a cylindrical co-ordinate system, but that's pretty
simple to do. And the fact that the control rods are
in the peripheries, we would need a transport in that
area, but we have that for MOX already, and that's
currently being tested.
So, we're using what we already have.
All right. So I'm going to summarize. I
think this was the slide that Professor Wallis has
been looking for.
CHAIRMAN WALLIS: So we can get close to
the end?
MS. UHLE: Yes. I don't know if it's the
end of my life or my career, or at least my
presentation.
The branch provides technical support to
the offices as needed, and we use the analytical tools
and, of course, the analyses capabilities of the
branch to meet those needs.
We're currently looking, the applications
we're looking at are associated with licensee
submittals, such as the power upgrades and the MOX
fuel. Generic issues such as the CRDMs, steam
generator tube integrity. Risk-informing activities,
50.46 and 50.61. And design certification, AP1000 and
pebble bed.
We realize that we will have to make to
improvements to these codes as emerging issues arise
and, again, we're focusing on doing that more in an
in-house fashion looking at perhaps coupling to other
codes as needed rather than using separate codes with
the same functionality. We're going to get away from
that. We're only going to use -- or only use what we
need to versus having ten codes in our code suite for
just TH.
And, of course, we're doing internal model
improvements such as for the AP1000 case.
Jack had talked about this, and I
mentioned it in the introduction, is that we have
hired -- we are in process of hiring entry level
employees as well to round out the technical
capabilities of the branch. Because we are actually
busier now than we have been in a while.
It's not just going to be for thermal-
hydraulics. It's also computation of fluid dynamics
as we start to use CFD more as a tool, especially with
the pebble bed work CFD will be used in the single
phased situations.
The severe accident in the fuel behavior,
we'll be ramping up the program and making a strong
connection in the branch so that we can work
seamlessly across the sections.
MEMBER FORD: Could I just ask a question?
Why entry level?
MS. UHLE: Because there's a lot of -- if
you look at the Office of Research, there's a lot of
experience in the Office of Research. And so with
staffing issues, we're not allowed to be top heavy and
all 15-10s there's some --
MEMBER FORD: Six to 1 ratio or something?
MS. UHLE: Yes. There's some, you know,
there has to be some ratio. And this idea of
everyone's going to start to retire, we need to bring
in entry level and mentor and, you know, have a more
gradual --
MEMBER FORD: I wasn't thinking of the 60
year olds, I was thinking of the experienced 40 year
old.
MS. UHLE: Experienced 40 year olds.
MEMBER FORD: Given the fact that you've
got a lot of workable --
MS. UHLE: We found some positions for 15s
in the branch that we are hiring in the severe
accident as well as the fuel behavior. In thermal
hydraulics, if you look at who has been hired, they've
been at the higher grade levels. So it's not all
entry levels.
I say entry levels, that's in some sense
we are more active in the entry level hiring because
there's more positions available. But the office is
looking at, you know, the higher grades as well.
MEMBER KRESS: Are any of those new hires
here?
MS. UHLE: Yes, they're all here. You
guys want to stand up.
MEMBER KRESS: They're all here.
MS. UHLE: Steve Bajorek. Da, da, da.
He's our SL, senior level scientist. He's our
experienced -- you're at least a 40 year old. Okay.
Joe Kelly, you know Joe Kelly. He is --
yes he's another -- he's another 29 year old.
MR. BOEHNERT: I think, Jennifer, you're
going to have to give your age now.
MS. UHLE: 32 November 23rd. I just turned
32.
MEMBER KRESS: You're not counted in the
new hires, are you?
MS. UHLE: What?
MEMBER KRESS: Are you one of the new
hires?
MS. UHLE: No. I've been demoted to
assistant branch chief. They won't let me touch the
code anymore.
CHAIRMAN WALLIS: Jennifer, we're way
behind in time. How long are you going to go on with
this?
MS. UHLE: Chris Murray. For
introductions, it's quick.
Chris Murray's from Penn State University.
Tony Ullses from NRR. He's been sparing with
Professor Schrock there for quite a bit.
And Joe Staudenmeier from NRR.
Chester Gingrich has been in severe
accidents. He was doing some thermal hydraulics work,
now he's going to go back to severe accidents.
And then, of course, there's Weidong Wang
in the back. Shanlai Lu and Jim Han is doing analysis
for us in the back. And Dave Bissette lead on the PTS
work.
CHAIRMAN WALLIS: We're severely behind in
time. Of course, you have given us more detail in
some of these things, since you were going to
summarize. Does that mean that we can move faster with
some of the later.
MS. UHLE: I think the question is how
many questions get asked.
CHAIRMAN WALLIS: Well, you had a
tremendous amount of stuff.
MS. UHLE: Well, when I went over it in my
head, it went very fast.
CHAIRMAN WALLIS: We need to be finished--
or you need to be finished by 1:30 because we have
another group, a very different group coming in and we
can't short change them. So we're going to take a
break now and then maybe you can work with your
colleagues to get us through on time. You work with
your colleagues to get us through on time.
And I'm a little nervous about Joe Kelly,
he always runs over. Maybe we could find a way to
prevent that happening.
So we'll take a break. Thank you very
much. And we'll start again at 20 to 11:00.
(Whereupon, at 10:25 a.m. off the record
until 10:40 a.m.)
MR. KELLY: My name is Joe Kelly, and I'll
be talking about the TRAC-M code consolidation and
development.
Now, the last time I was in front of this
Subcommittee I was up here for 6« hours. And since
we're already an hour behind schedule, Professor
Wallis is concern is well taken.
So this presentation really is three
presentations in one. I was going to talk about the
code consolidation status followed by Jennifer Uhle
talking about the SNAP development that's the
graphical user interface. Then I was going to talk
more about our long term development plans and a
movement about how we're going to integrate some of
our stand alone programs into the code development.
So what I'm going to do is condense the
code consolidation status, you've heard a lot of this,
in half, and Jennifer is going to skip this
presentation, because you've hard about SNAP before,
and then I'll try to give most of what I had planned
to give.
When we first started this program back
almost 5 years ago, we laid out five areas that we
wanted to make improvements in. Modernize the
architecture, accomplish the code consolidation to
conserve resources, improve the ease of use, accuracy
and numerics. And I was going to say something in
each of those areas, but I'm going to shorten it
because first I what wanted to do is give you an idea
of where we are today. And that should somehow avoid
--
Take my word for it. The colors on the
view graph are much prettier than the colors here.
That's really horrendous.
But anyway, this is where we are today.
We have accomplished the modernization and the
functionality. We have parts and functionality of
TRAC-B and RELAP5. We do not have physical models of
those codes, nor do we intend to implement all of the
physical models of those codes.
What we're working on at the moment is
called the component mapping, and that's the way that
you take your RELAP5 component through the SNAP
graphical interface and translate it to a TRAC-M
component. And that's what's going to enable us to
take the RELAP5 input deck, read it in and run it with
the TRAC-M code. This work is almost complete. This
line is supposed to show about where we are. It will
be complete shortly after the beginning of the year,
at which point we'll start a development assessment.
Now, originally the idea was to start the
assessment and let the model deficiencies from TRAC-M
show up as a result of the assessment. Then when you
identify a deficiency, go look at them. First, go
look at the models in the other code and try and make
a judgment that, say, interfacial drag in TRAC-B is
better than TRAC-P, etcetera, and then bring that
model in. And that would then be a cyclical process.
Now, we're still going to do that to some
extent, however there are two deficiencies that
immediately showed up. The first is rod bundle
interfacial drag, and that's what we've alluded to in
the Peach Bottom Turbine Trip when Jennifer was
talking about that. We simply couldn't predict the
action in a void track and operate the BWR accurately
enough. So what we're going to do is implement,
again, basically the interfacial drag and interfacial
heat transfer routines from the TRAC-B code to be used
only for BWR channels and probably also the BWR core,
but not globally.
Likewise, this is a deficiency that has
been identified in the reflood model. I'll talk a
little bit more about that. That's what I'm working
on.
These will feed in as soon as they're
finished through developmental assessment, and we'll
have roughly about a six month period where all the
models of the code will be frozen and go through the
entire assessment matrix and then that leads to
releasing the consolidated code at the end of calendar
'02.
CHAIRMAN WALLIS: So maybe by the middle
of next year or something you can show us some of your
development assessment work?
MR. KELLY: Yes.
These are the type of slides I'm not going
to belabor. The only thing I want to point out on
this one is that we have something called an exterior
communication interface, and that was built in to
allow us to very easily couple the other codes or
special modules into the codes for capabilities that
we don't either have in TRAC-M or don't want to build
in. It's already been done in an explicit with the
REMIX code, the PPS calculations, and also we've done
a preliminary coupling with the CONTAIN code.
We're skipping the SNAP presentation, and
I'm going to not belabor this also, but we've put a
lot of work in the draft communication interface
making it easier to use. So if you ask our new group
what is their highest priority item, this is it.
Most of what we need is going to be done
in early 2002, but the playback capability will be
mid-2002 and interactive display with user feedback,
that is where you can run it like a simulator mode, is
sometime in the future.
Documentation was mentioned earlier.
Documentation is a very important step and it has to
be a continuing effort over the life of this project.
CHAIRMAN WALLIS: Is it true that the code
has not yet run?
MR. KELLY: Excuse me?
CHAIRMAN WALLIS: The question I get is
that it hasn't yet run, because it hasn't yet done
these PWR transients or PWR LOCA or anything?
MR. KELLY: No, we did those. No. The
code runs and it has been all throughout the process.
We did the development that way. And there are
several hundred test problems designed with each
developmental version.
I shouldn't have skipped probably over
this. It's seeing results of TRAC-M coupled to PARCS
for the Beach Bottom Turbine Trip as well as a main
steam line threat. So that is TRAC-M doing those
calculations.
We can do BWR to our transients now, the
reason I say early 2002 here is so that we can read in
a TRAC-B input deck, and existing one, and run it in
TRAC-M. All that capability is there. But the reason
it has this date on it is for the upgrade to the
interfacial drag package. Which Tony Ullses is trying
to quickly put that in to see if it would work and
make the improvements that he needed for Peach Bottom,
but we want to put it in a more correct way according
to what we call a low level modularity. And so that's
when this work will be finished.
For the SBLOCA, I don't know that we've
actually run any of those. The completion date here,
though, is for the component mapping, you know, when
that development work will be finished. And that's
when the assessment for SBLOCA applications will
start.
For large break LOCA, we could do a large
break LOCA now but from my standpoint the reflood
model was flawed so that this is the date by which
we'll have an interim reflood model and we'll start
doing the reflood part of the assessment matrix.
MS. UHLE: Joe, can I just clarify one
thing on that. I'll only be a second.
This is the last --
MR. KELLY: Just don't get my slides out
of order.
MS. UHLE: I know. I am just going to --
this was going to be my presentation on the buoy.
Again, here, with respect to RELAP release, we have a
RELAP5 version completely finished for the post
processing and the model editor where you're dragging
and dropping models. We can interact with that
display. You want it in a simulator mode already with
RELAP and TRAC. The date here being future is that
with the idea of having -- we have a three dimensional
model and right now when you look at the playback,
you're seeing it in 2-D. We want, for the ease of use
for the user, extend that so that you can represent
the three dimensionality in a more easier way. So
that's why under this there's a future, although we do
have the ability. We showed you that last year,
running of TRAC in an interactive mode while we opened
valve and saw it blow down. So that's been in for two
years -- I mean, for a year.
And in early 2002 being able to run,
taking a RELAP5 input deck and converting it to TRAC
and doing the drag and drop through the TRAC model,
that's the last bit that we're doing right now. And
also the plotting here with the mid-2002 date. Again,
that's associated with three dimensionality. We can
2-D plot now already. We want to be able to 3-D plots
very easily to get the surface plot of the core boil
fraction and the 3-D kinetics.
MR. KELLY: Good. Thank you, Jennifer.
CHAIRMAN WALLIS: I notice the
documentation is a continuing effort.
MR. KELLY: Yes.
CHAIRMAN WALLIS: Doesn't the
documentation come first or do you write the code and
then figure out what you did and write up the
documentation?
MR. KELLY: As Jennifer said earlier, we
have conformed with a fairly rigid SQA, certainly
compared to anything that's ever been done with NRC
code.
CHAIRMAN WALLIS: Then this documentation
should be in good shape.
MR. KELLY: Yes, but each piece, each new
piece has to be folded in in the overall
documentation. And, for example, it was mentioned
earlier that the TRAC manuals were extraordinary and
it's hard to find your way around in some of them.
Rewriting all of that from scratch is a major task.
And what we're doing at the moment is basically
putting in the pieces that we're changing. We do need
to go and make all the whole restructure done but
that's a huge effort and we've connected randomly to
make it work at the moment, but we are going to --
that's why somebody has to start and keep working at
it.
CHAIRMAN WALLIS: But it's important. The
way you present the documentation is important; that's
what's out there, people look at it.
MR. KELLY: Yes.
CHAIRMAN WALLIS: It's got to be credible
and not have typos and all the usual stuff.
MR. KELLY: Right.
MR. BOEHNERT: Historically what's
happened is the documentation was always put off the
end and then somehow it never got done.
MR. KELLY: Right. What we're trying to
do is have the people as they develop a model or
implement a component do the documentation for that as
part of the SQA. But it's still does need to get
folded in better to a master document. We're not
there on that yet.
CHAIRMAN WALLIS: As I said earlier, we
can help in the early reviews of this documentation.
MR. KELLY: That would be very good.
CHAIRMAN WALLIS: We'd like to do so.
MR. KELLY: We also need to, as you know,
improve the code accuracy. And really what I wanted
to say here is we're beginning now to put the models
in the code. And that's a huge effort, but what we
focused on for the last few years is putting in
capabilities of the functionality consolidating. But
we've got a lot to do here and this is just starting.
But that's not part of the code consolidation, that's
part of us evolving to this actual state-of-the-art
thermal hydraulic code.
CHAIRMAN WALLIS: Beginning isn't a good
word, though.
MEMBER SCHROCK: One of the problems has
been that the codes run part of the way through a
problem and then crash, and then people fix it up and
run the rest of it. In my mind that leaves a lower
level of reliability when that kind of thing happens.
Do you have an objective for this code
that that is not going to be allowed or is this going
to be a continuing problem?
MR. KELLY: What we have the objective of
is to improve the robustness of the code and that is
just what you are talking about. It is making the
code be able to run to completion and not only run to
completion, but run without these periods where it
just grinds to near like halt and you go to, you know,
10-6 time steps.
So what we're going to do is starting in
the assessment when we start running in to those
problem, the code either fails or it has significant
swim outs, we're going to, in effect, ship that
problem off to our numerics guru, otherwise known as
Professor John Mahafty, who is going to help us track
down the root cause of it.
Sometimes it'll be the numerics. You
know, some like the way the water tracking interacts
with level tracking of whatever. Sometimes they will
just have to be an old condition numerical, an old
condition physical model, the physical model that
causes, you know, oscillations or causes you to
accelerate your condensation as you go to saturation,
which makes it hard to put numerics to solve. In
which case if it's a physical model, John will kick it
back to me and we'll work together to try to make it
more robust. But, again, that's going to be a
process. It's going to be a process over a lifetime
of the code. But it is something we're committed to
provide. John?
MR. MAHAFTY: Yes, this is John Mahafty
from Penn State.
If I could make one comment on that. You
know, I've given guidelines to people at NRC and other
places that if the time step gives 10-5, there's
something wrong with the code, and I should see it.
If it runs for any significant period of time below
10-4 there's something wrong with the code and I
should see it. So, you know, we're not taking the
kinds of shortcuts -- that's a good term -- that were
done in the past. And I've seen problems where people
have run RELAP5 and it grinds down and runs at 10-6
seconds for long periods of time and finally it
recovers and goes on. That's not acceptable for us,
because it tells you there's something wrong with the
code, some kind of numerical problem is potentially
masking what physically should be done and audited.
It needs to be looked at and it needs to be fixed.
MR. KELLY: And I agree completely.
The numerics can also effect accuracy, and
there are a few things here. In the future we'll be
looking at higher order differencing in order to
resolve things like thermal fronts. As most of you
know, the difference in the code at the moment is
first order accurate upland differencing so it tends
to smear out sharp interfaces. Thank will be future
activity. One that we have gotten created is level
tracking. And level tracking doesn't just mean, you
know, we are a 2 face interface hits and where is this
continuity of void fractions. What it means is you
find where that is and you go in and modify as part of
the time step your mass energy and momentum
conservation equations to take account of where that
interface is on your computational grid. And I'm
going to show you an example in an oscillating
manometer problem of why that's important.
In this last, we reimplemented semi-
implicit scheming code which turned out to be very
revealing in order for that we could do our stability
calculations so that you don't get the damping that
can develop in implicit scheming.
This is an oscillating manometer test
problem. Very simple. Two vertical pipes, they're
each 10 nodes one meter long. And this is collapsed
liquid level versus time. The two pipes are joined at
the bottom, they're open to the atmosphere at the top
so it's an air-water simulation. They were
initialized half full at the 5 meter level with a
velocity such that this should oscillate with an
amplitude of 3 meters.
CHAIRMAN WALLIS: No friction?
MS. UHLE: No friction.
MR. KELLY: No friction. Water pressure
is turned off. Thank you.
Hence, we can tolerate it, but the orange
curve is an analytical solution, and when this was cut
and pasted from the frame maker document into
PowerPoint the curves got kind of shaky. But this is
an analytical solution showing no dissipation.
The black curve was the TRAC-M calculation
with a standard curve. And after about two cycles it's
totally damped out. And the reason for that has to do
with the discretization of the momentum-flux terms
across that sharp interface.
CHAIRMAN WALLIS: It's a numerical
diffusion, in a way.
MR. KELLY: Yes. It's an artificial --
CHAIRMAN WALLIS: Artificial --
MR. KELLY: -- viscosity that wasn't
intended, but because of the way the two phase
momentum-flux changes. When you correct that, and
this was work done by Birol Aktas of ISL, this is what
you get.
Now, the test problem was changed slightly
--
CHAIRMAN WALLIS: Are all those points on
the curve, those are predictions?
MR. KELLY: Yes.
CHAIRMAN WALLIS: That there on a big Sine
wave which is in length about ten times the --
MR. KELLY: Okay. The legend is missing
here. The upside down triangle is simply an
identifier for the curve. It's not a point.
CHAIRMAN WALLIS: Oh.
MR. KELLY: And likewise, so what you're
seeing are two curves sitting right on top of each
other. And they're virtually indistinguishable, which
is very nice that we can actually reproduce the
innerlocal solution. But not only that, we make the
test problem a little bit more difficult. It still is
two pipes, but it's actually now six individual pipe
components so that we could make sure that the level
traction in steam could cross boundaries between pipes
smoothly without putting any dissipation between that.
So, as far as the level tracking concern,
there's no difference now between a no boundary in a
pipe and a boundary between pipes. And it's just part
of the QA process to make sure the model works.
There have been a number of improvements
to the --
CHAIRMAN WALLIS: It seems to me there's
a whole slew of these QA models you need to check, not
just this one.
MR. KELLY: Right. And the more of that
we can do the better.
CHAIRMAN WALLIS: And I think it's been
one of the concerns with all these codes that they're
okay for nuclear safety, but they can't predict some
of these very simple lab experiments.
MR. KELLY: Yes. I don't hold to that
theory. I think you have to predict the phenomena
that are actually there.
CHAIRMAN WALLIS: I think you should,
right. It's got to be honest.
MR. KELLY: And that's going to be a
process, and one of the most important parts of this
program is going to be the assessment. And that's got
to be a continuing activity at a fairly high level for
years, and just continue.
CHAIRMAN WALLIS: I hope you keep doing it
already. Have been doing it.
MR. KELLY: I'm not going to really talk
about the improvements to the kinetic module. That
was pretty much gone over in Jennifer's presentation.
I simply don't have the moxy to do it.
MEMBER SCHROCK: What did you do to the
numerics again to change the picture so drastically?
MR. KELLY: Okay. If you get me off into
details, I may have to go to Birol, but I think I can
give you the idea.
MS. UHLE: Birol left. John Mahafty is
his thesis advisor, he can answer the question.
MR. KELLY: If you have an -- in the
momentum-flux term there's an alpha row of DVDX. How
do you discrotize that term across an interface is the
problem. And if you look at the way it's typically
done in RELAP or TRAC normally, it's really built into
the two fluid model an assumption that you have these
continuous evolution of weight fraction across the
computational mesh. And when you do that, that term
is suitably accurate. But if instead you actually
have a sharp interface, so let's say you're on
convection vapor out as this interface goes, but
you're averaging between these cells to get these
alpha rows and DVDXz then you introduce a dissipation
term.
And I've actually even seen in some codes
when it said dissipation, in fact sitation will reduce
oscillation. But normally it's dissipating.
So what we've done is say we have this
tracking scheme that tells where this level is. Now
in our, you know -- we basically pull our back of the
envelop and write down what the momentum equation
should be if you've got the single phase vapor going
across this with this level approaching it. And then
put in, adjust the terms in the momentum equation and
make them what they really should be.
MEMBER SCHROCK: But I thought there was
already a level tracking in the original track.
MR. KELLY: There was one in TRAC-B.
MEMBER SCHROCK: Yes.
MR. KELLY: And there was -- it works more
as an interface sharpener. So what it would do is try
to track where the level is and adjust interfacial
drag in an interfacial heat transfer model. Level
tracking has to, if it's going to work right, has to
do a lot of things. And so it basically it turns
interfacial drag down. It says, okay, there should be
a level here. Let's lessen interfacial drag so we
don't pull this liquid up when we shouldn't be.
Likewise, it says okay the interfacial area is a pipe
instead of treating the vapor as bolts.
But that's just really -- that's the easy
part. The tough part, which really gets this to work,
is going in and actually fixing the conservation
equation for a different physical situation.
CHAIRMAN WALLIS: So this goes back to
what I said this morning. Jennifer was talking about
a pipe. You can recommend an equation for a pipe, and
in fact they were this way, so that it behaves like a
pump.
MR. KELLY: True.
CHAIRMAN WALLIS: Under some circumstances
because of the way you're averaging the stuff.
MR. KELLY: You have to be very, very
careful. And this is something that Birol under
John's guidance did a very good job on.
The last stage of the consolidation
program is developmental assessment. And what I've
done is put together an assessment matrix that we're
going to start to do during calendar year 2002. I'm
going to give you an example of how that was put
together.
I also have a handout, I'll give you what
I've proposed test matrix says. I've got that written
in, I'll get it to you.
Now, the test matrix is quite extensive,
but it is far, far from comprehensive. I mean, there
are whole areas that are left out. And those areas
are going to have to be plugged by the assessment we
do in the future. And that's why Steve Bajorek is
going to talk after me.
We're going to a per face developmental
assessment for each of the applications that the code
is going to be used for.
So what I'm doing here is, remember our
success criteria for the consolidation. For the TRAC-
M code we will be able to run it against each of the
predecessor codes; TRAC-B, TRAC-P and the RELAP5 for
the application of interest for each of those codes,
and TRAC-M would do at least as well. That's our
success criteria.
CHAIRMAN WALLIS: Do you have a matrix
like this for simple experiments, like the manometer
as well as these --
MR. KELLY: Yes.
CHAIRMAN WALLIS: -- messy experiments
where everything's going on and you get AB
compensating errors and so on?
MR. KELLY: There's about half of those in
the works. And that's something that could be
expanded.
This one is for separate tech specs
reflood heat transfer. And what I'm going to do is
just give you an example of how this got made up.
The first thing I did was for the three
predecessor codes; this was TRAC-M the F77 version,
which is basically just TRAC-P. There are no models
in this version at all. An assessment of that was
done relatively recently, and that's the document
NUREG/CR-6730, and that was published, I think, about
a year ago.
For TRAC-B the last NUREG-B developmental
assessment code was, i believe, the 3663, and after
that there were two other NUREGs by other contractors
that did further assessment of the TRAC-B code. And
I also had input from INEL and Penn State.
For RELAP5 this was the last published
development assessment of the code, because there was
also an assessment of it in this NUREG as well as the
assessment we did for as part of the AP600.
So I looked at all of the tests that were
done for these, and for this phenomena listed each of
the ones according to the code it was used for. And
we then in the TRAC-M column, I basically summed them.
Now, if we simulated with one of the other
codes, I brought it over and stuck it in here. And so
these are the ones we're going to do unless there was
some reason not to do so, and that logic is what I'm
going to show you now.
All of the codes the flux is at 31504;
that's a one inch per second 40 psi base case force
reflood test. So, obviously, we're going to do that.
31701 is 6 inches a second, so that's at
the other end of the spectrum, so that was done in
RELAP5 and we will include that here.
Now, this test 33436 is a gravity reflood
test done in FLECHT SEASET, and because of the way the
downcomer is and the way the exit coming up the front
is there are a lot of uncertainties in the downward
positions. So there's no good reason to do a gravity
reflood simulation for that facility when we have
facilities like CCTF and SCTF. So I'm going to
eliminate this test.
Now, when I looked at -- I just mentioned
CCTF. As part of the TRAC-B assessment measure there
was a CCTF basis run in 14. But what they did is
actually a gravity test, but they ran it as a forced
reflood test. What that means is they stripped off
the downcomer, stripped off the wall clamp and imposed
a flooding rate at the bottom of the core. That's
artificial. No one knows what that flooding rate is.
They inferred it, it was inferred from the
experimental data based upon what came out the top of
the bundle and the build up of inventory in the
bundle.
So running this as a forced test -- I mean
if you're not even monitoring what you're putting into
the bundle right, how can you do an assessment of it.
So I think this is of limited value and as part of the
integral effects testing we will be doing a couple of
CCTF cases. That's something in the future we'll have
to expand. I saw no point in doing this, so I took it
out.
FLECHT SEASET there was one test run for
TRAC-B, and obviously we're going to keep this. This
is a large scale reflood gravity. It's 8 bundles,
2,000 meter rods lined up in a slab, so it models at
full scale the distance between the reactor core
center line and the core barrel. Very important
contributing effects. The Lehigh Rod Bundle, this was
done with TRAC. It's a nine live bundle, so it's 3 by
3, which means it's about this big, and there's a
heated shroud there's a lot of questions in its regard
to things like heat losses, its quality fully
instrumented and plus as you know, if you try to do a
two phrase, and this is also in one atmosphere, test
in something this big, any vapor structure is going to
span it and it's going to act not like a broad bundle
at all, but more like a tube. So it's not productive,
its of limited usefulness, let's not waste our time on
it.
FLECHT test 9077 which was done on TRAC-B, is
from the original FLECHT series and it's 6 inch per
second new core rate capacity. Now that facility does
not have delta P cells, and likewise did not measure
specifically the steam temperatures. There is a lot
in that less experimental information than there is
with the more modern codes like FLECHT SEASET. So I'm
going to get rid of this test and replace it with
31701.
The GOTA, and I know I don't pronounce
that right, but this reflood test is combined top
spray cooling and bottom reflooding. Well if we're
going to discuss the BWRs you need to have assessment
cases for that, so we've got to keep this.
The NEPTUN facility which was done in
Switzerland is 33 rods of half height. Now, again, 33
rods as counted as 6 by 6 with the corners taken off,
is relatively small.
The two tests that were done, one was at
one 1« centimeters and one was at 15 centimeters a
second. So what I've done is instead of doing these
two tests, I'm going to substitute the FLECHT SEASET
test run. This one is 6 inches a second. I added
this test, which is 34006, which is 0.6 inches a
second to compensate for the NEPTUN test that I'm
going to drop out. So what I'm trying to do is to
come up with a test basis that makes sense and covers
the range of conditions that we have been testing
before.
CHAIRMAN WALLIS: Of course, the advantage
of something like NEPTUN in is that it's not -- your
conclusions are not test dependent so much on FLECHT.
You can say you've got something independent, you're
able to predict. And if there's something wrong with
the modeling because of the geometry of NEPTUN, maybe
that means that something should be in the code
anyway. So you might see if you can get a more
diversity, perhaps, in the sources of the experiments.
MR. KELLY: What we will be doing --
remember, these are -- we will be doing in the future,
okay. We're going to expand the matrix both in the
CCTF and SCTF, and this is something that Steve is
going to talk about. There are actually forced
reflood tests in SCTF, which are of great value
because then you know actually what you're showing in.
You don't have the complications of, you know,
potential oscillation and some down time.
CHAIRMAN WALLIS: You've got to prepare
NEPTUN. When you sort of release the code somebody
else may, and you might want to do it ahead of time.
MR. KELLY: The problem is there's a lot
of data out there and we have to just do the best we
can.
MS. UHLE: Our international partners
have, especially Switzerland with respect to NEPTUN,
are interested in doing assessment for us. And that's
what helps us get, you know, broadening our assessment
range. And so I think that first to meet the 2002
deadline, we are trying to make one that, you know,
take some consideration with the good data that's out
there and then in the future, with the fact that we
have the PM program, that really broadens out our
assessment on this.
MR. BAJOREK: We would welcome other
groups coming in, any additional tests that would have
matrix. One of our problems has been resources and
trying to pick the test step that will give us the
most information without letting the matrix get to out
of hand.
MR. KELLY: And I'll give you a copy of
the proposed matrix as soon as I get off the stage
here.
That ended the part the presentation on
the status of the code consolidation. And we're going
to skip over the presentation on the SNAP, and what
I'm going to jump into now is instead the code
development effort for the future.
Again, when we first started this project
we went out and queried our users, both internal and
that would have been NRR and was then a ADOD, as well
as PT and RAS. And our external users liked, you
know, he said, if we're going to have a state of the
art thermal-hydraulics code, what should it have in
it? And this is the laundry list they came up with.
In items number 1 was an improved user
interface. And that part of the reason why we're
putting in the effort on this now, to make the code
easier to use.
I'm not going to go through all of these,
but I'm going to do instead --
CHAIRMAN WALLIS: Well, where there's a
gap, does that mean you're not doing it at all?
MR. KELLY: No. It means that that it
hasn't started.
And what I'm going to show you now --
CHAIRMAN WALLIS: You're not using modern
numerical method?
MR. KELLY: Well, that means that
developmental efforts incorporate, for example, either
higher order differencing or a more fully implicit
scheme has not started. And for what I'm going to
show you, again the colors are abominable on the
viewgraph. I don't know what it worked out that way,
but --
MS. UHLE: They're extraordinary.
MR. KELLY: To say the least. This is our
plan for what we're going to do next year and the in
the future.
So, up to here is the conclusion of our
current five year plan. From this line on is the
future.
Now, I've broken this down into these
categories: Consolidation and assessment; physical
models; numerics improvements; modeling capabilities;
and, then along the bottom I've shown code release
dates. And this Rev zero will be the first release of
the consolidated code, and that is at the end of 2002.
And what we're planning is annually at the end of each
calendar year to release another revision to the code.
And now let me explain this a little. If
you could see the colors, there's a color code here.
This is supposed to be a light blue. You notice these
boxes go with this code release. So these activities
will be finished and go into this code release.
Likewise, the green boxes feed into this
one.
CHAIRMAN WALLIS: You must have had a
color consultant or something.
MR. KELLY: Well, apparently I didn't have
a very good one. A budget decrease.
And so forth. And then actually this is
revisions 3 through 5. I didn't -- once we get out
this far in the future I don't know exactly what we're
going to be doing when.
CHAIRMAN WALLIS: Why didn't you use the
primary colors?
MEMBER SCHROCK: If you put actual dates
in there, whatever they're going to be, 0.0 is
October 1, 2002.
MR. KELLY: No, these are the calendar
year.
MEMBER SCHROCK: Calendar year.
MR. KELLY: Yes.
MEMBER SCHROCK: So that means January 1,
2002?
MR. KELLY: Exactly. That'll be the first
release of the consolidated code.
MEMBER SCHROCK: So that's 13 months away
and you don't have a document that describes the code
in any complete way today. You intend to have one
prior to that and have some feedback as to how good it
is?
MS. UHLE: You want me to answer?
MR. KELLY: Please.
MS. UHLE: We have -- I mean, because we
started from a TRAC-P code, we have the base TRAC-P
theory manual. And in-house we're going through that,
Frank Odar, Jim Han and they're pointing out where
things are confusing. With the developmental work
that has been going on we follow S2A procedures. And
when it involves physical modeling, of course, there
are sections written by the developers documenting
what was done so that those sections will be put into
the theory manual. So we have the documentation. It
has to be merged and it has to, again, get another
read through to make sure that there are --
MEMBER SCHROCK: I guess what I'm asking
is are you going to release this whether the
documentation has been reviewed or not?
MS. UHLE: No, no, no. We will have the
documentation released with in-house review and if
you're offering review from the ACRS if that's what
you're offering. But, yes, I mean we realize that
it's fast approaching. But I don't want you to think
that there is no documentation.
If you go up to the consolidation room,
there's documentation like up to here. It's a matter
of going through, organizing it and putting it into
the master document. Now everything is written in the
same way, word processor format, so that's going to
facilitate things. And then we're starting to begin
the merging.
The user manual is up to date. We have to
put in modeling approaches on how to model the BWR,
but we're going to take out of the TRAC-B and, again,
go through read through and add to it as necessary.
But the user guide is the one that's in the best
shape, and the theory manual is our one that --
especially during the developmental assessment and we
start to replace physical models, we'll be adding to
that.
The programmer's guide talking about the
architecture of the code, we have made revisions to
that with the modernization, but that one is the one
that's lagging the most, although because we're
focusing on making this code more maintainable that is
something that we will have to do.
MR. BOEHNERT: Does the master document
include those three things you just mentioned?
MS. UHLE: No. There's a theory manual,
that is a master document. There's a user guide, that
is a master document.
MR. BOEHNERT: So each one -- okay.
MS. UHLE: But, again, if you're
interested in looking at what we've generated so far
and reviewing it, then we would be, I would think,
more than willing. Although I don't know, I'm not the
office director.
MR. KELLY: Thank you, Jennifer.
CHAIRMAN WALLIS: Is this thing suitable
for any two phase flow problem? I mean, it doesn't
have to be nuclear reactors, does it?
MR. KELLY: Well, you can keep the
components of a couple of different pieces.
CHAIRMAN WALLIS: I think it would add a
tremendous amount of credibility if it was something
like the commercial code which is out there and has
been proved to work for oil and gas, and chemical
plants and all kinds of things. If it works for all
these other areas as well, then it must be really
good. When it's only been shown for a couple of
nuclear applications, then it looks real suspicious.
MS. UHLE: We are getting requests for the
code for the oil industry. And also heat exchanger.
CHAIRMAN WALLIS: It would be very nice if
you could in some of these presentations, particularly
the public presentations show that it's not just been
tuned to some nuclear applications. Okay.
MR. KELLY: For the consolidated code,
what is going to show up is what we've talked about
before, finishing the way to translate RELAP through
SNAP to be able to run it in TRAC-M, the developmental
assessment and then there's the two model changes, the
bundle interfacial drag, which is an implementation of
the TRAC-B and the interim reflood model, which I'm
working on. And then we are going to do some work in
the beginning of the year on robustness.
And one of the things I tried when I set
this up was have development activities in
approximately in mid-year so that we would have a
frozen type version for, hopefully, as much as six
months to go through the testing before you get to the
release date. So that on December 31st we're not
changing the code model of a code that we're going to
release January 1.
Probably one of the most important
activities here is the PIRT based assessment which I
show across the top. And that's what Steve is going
to talk about. And it can be this assessment where
you look at the important phenomena and see how well
the code does against them that then will drive what
we do here.
The only other thing I want to talk about
is some of the model development from out experimental
programs. The green box here is supposedly subcool
boiling, and that refers to the UCLA program on
subcool boiling and low pressure. We're going to
receive a model approximately mid-year and we'll be
implementing it during the end of 2002. But because
it's going to come in at the end of year, I don't want
it in the release code version because we want in for
suitable testing. So it will be part of the Rev 1.0
release.
This box is phase separation and this is
to build on the experimental work at OSU. So when
we're able to get a model from them that we have
confidence in we'll be putting in the code, hopefully,
in early to mid-2003 to show up in the Rev 1.0 code.
The other one is -- it really should be
more mechanistic but obviously we're not thinking we
put first principles, but more mechanistic than
certainly what we have today. And that's going to
build on the external information from rod bundle heat
transfer facility at Penn State. And I have several
slides on that later.
These tests are not yet defined, but I
have an idea of what's going to go in them. And those
actually take the next few slides. I'm not going to go
over those in detail, but this is what we anticipate
as of today that we're going to have change in the
code to make it really do a good job on more and more
reactors. Core spray model, boiling transition. For
example, the model incurred is normally the OXY
correlation, which is basically the annular pore
regime in tubes. Obviously that does not give a very
good representation of dry auditing of water reacting
models, but also putting the place in the code that
supports, if you will, where a user in NRR or actually
at the request of NLR can incorporate on a temporary
basis a proprietary model in order to help them
facilitate their review. You have to adjust modern
fuel designs and fuel designs. And obviously, as I'll
show you, the reflood model needs a lot of work. That
applies to more and more reactors as well as to
pressurized water reactors. And also we'll have to
look a little more at top-down rewet both on the
channel box and the fuel lines.
MEMBER SCHROCK: This item on the BWRs
incorporated the proprietary model, you have in mind
something like what GE says they have for their rod
bundle on the first principle.
MR. KELLY: I assume what you mean is
where they have a drop of pH and look at the stripping
of the drops and the deposition of the drops
downstream of the grid. That's not what I meant.
What I meant here was the better correlation. Like,
for example, I'm looking at pressurized water
reactors. Each vendor of each fuel design goes
through a testing program and develops and in effect
licenses the correlation for that particular type of
fuel. And then if you go and do some kind of
operational transient where your success criteria is
DMBR margin, well if you forget to write thermal-
hydraulic conditions versus time, but you want to
check the margin, you need to have an actual
correlation for the BMBR that suits that fuel
geometry. And it even depends upon all the little
tabs on the rib spacer, and it's somewhat analogous
for boiling-water reactors.
MEMBER SCHROCK: Well, here's you're
talking about boiling-water reactors.
MR. KELLY: Right, that's true.
MS. UHLE: Can you do a Drexal
correlation?
MR. KELLY: Well, what I'm talking about
is more like that.
MEMBER SCHROCK: That's what I'm talking
about.
MR. KELLY: Not going to actually trying
to predict it by stripping the film off the rods and
then depositing the drops downstream; that would be a
research project.
CHAIRMAN WALLIS: Joe, you are half way
through your slides and taken about the time that was
promised.
MR. KELLY: Really? I thought this was
going pretty fast. Okay.
MEMBER SCHROCK: One last simple question.
Do you envision this option to incorporate the
proprietary model as something to be used by industry
in their use of the code or something you would do
with your code?
MR. KELLY: I envision it as something
that we would do in order to facilitate doing our in-
house calculations. But it would be something that
other people could use to more easily implement.
MEMBER SCHROCK: Yes.
MR. KELLY: And this is, you know, just
what we would like to do.
MR. BOEHNERT: Well, I don't know how you
get around, though, the thing that these codes are
supposed to be publicly available. I mean, that's --
MR. KELLY: That's why we're not going to
build it into the code.
MR. BOEHNERT: Yes, I understand.
MR. KELLY: Just a box, you code it
yourself.
MR. BOEHNERT: Yes, a black box you put it
in. Yes. Okay.
MR. KELLY: Similarly, I've looked at, you
know, based on what we've done in the past, we looked
at small break LOCA, what were the problem areas. And
we made up a laundry list of where I think once we
really start doing the PIRT based PA we're going to
have problems. And this is the list, and in the
essence of time, I won't go through the list.
I have a similar one for large break LOCA.
And that takes me to what we're doing now, which is
the current model development activities. And there
are two, as I've mentioned. The first is not model
development so much as it model implementation, so
it's a rod bundle on interfacial drag, boundaries
necessary for the Peach Bottom Turbine Trip benchmark.
What we're going to do is implement the TRAC-B
interfacial drag and interfacial heat transfer models
all in route for the CHAN which is a BWR fuel
assembly. And we're going to look at applying it to
the core region of the 3-D vessel. Because,
obviously, the interfacial drag per bundle is better
than the correlations we have at the moment, which
were mainly focused for 2-D. And it'll just be
implementing them at what I call low-level
modularization. This is an in-house effort by Joe
Staudenmeier and Tony Ullses.
The development activity is to come up
with an interim reflood model, and it's necessary for
doing realistic auditing calculations for the AP1000.
The reason it's necessary is the current model has
unacceptably large oscillations and at least for
separate FLECHTs tests it's highly conservative. I'll
briefly show you the results of one of those.
We have to look at two things; the
physical models and also the fine-mesh numerical
scheme, and also is an in-house effort with Weidong
Wang and myself.
I'm going to skip the fine-mesh rezoning
scheme, just in the interest of time, unless there are
questions about it.
So I'm going to skip over the next two
slides.
MEMBER SCHROCK: You never question the
adequacy of flow regime maps in the code.
MR. KELLY: Well, I do. Do you mean the
idea of using flow regimes in general or the ones in
the code in particular?
MEMBER SCHROCK: Well, I mean the ones in
the code in particular.
MR. KELLY: Yes. I mean, certainly
something that's based on a one inch diameter air-
water atmospheric pressure is not anything close to
what, you know, we should be having in reality. And
that's something we have to look at. There's lots of
areas of physical models --
MEMBER SCHROCK: But it isn't going to be
a part of the TRAC-M development?
MR. KELLY: Not part of the development to
be released at the end of December 2002. In my
master--
MEMBER SCHROCK: So you think it will be
eventually?
MR. KELLY: Yes.
MEMBER SCHROCK: Okay.
MR. KELLY: I'm pretty sure. There was an
item on here for low pressure interfacial drag that I
didn't talk about, and we're pretty sure that once we
start doing things like AP1000 and low pressure EKD
models that we're going to over predict interfacial
drag. And that's a point where we revisit the bundle
interfacial drag model and try to establish a database
and maybe come up with a new model if we can't find or
develop one that is accurate enough.
The whole idea of replacing flow regimes
is out here much later in time, and that's the
interfacial area transport work.
But speaking of flow regimes, there are a
number of idealized points in reflood. And what I'm
showing here is clad temperature versus time of 1 inch
porous reflood case, and this is the heat transfer
coefficient versus time. And so at any one point you
do through a progressional regime starting with steam
cooling. The steam cooling actually probably stopped
here, and this is when the dispersed flow film boiling
started. The dispersed flow film boiling started as
the most important regime simply because that's the
point in which the turnaround in the clad temperature
established a peak value. So you always think you
need to model this very well. However, there's
another regime just a little up stream of it which
I've labeled the froth region here. And in the future
you'll hear me talk about invert slug, invert annular,
those types of things.
This region could be anything from a few
inches to a couple of feet, depending on the flooding
rate and liquid subcooling. It's very important from
the standpoint that with this cooling that brings the
clad temperature down to your quench, the temperature
at which the rods get wet. So not only does it
control the propagation of the quench time, but it's
again that the vapor generation in this area and at
the quench time provide the source term for vapor flow
and entrainment that you have in the dispersement
area.
Currently in TRAC-P -- this is from the
manual, this is the reflood heat transfer coefficient
module. And I don't expect you to be able to read
that from this, but it's okay.
This is an imagine the idealized flow
regimes when we go from transition boiling, smooth
inverted annular, rough wavy, agitated inverted
annular, dispersed flow, highly dispersed. In all
these different regimes, you go through the code and
you use a weighted sum of contribution to each regime.
So what you have is one model turning on, ramping off
and another model turning on and ramping off and so
on, and you add all these pieces together. Well, it's
highly confusing, it's also very complicated.
CHAIRMAN WALLIS: That is the problem in
using a high pressure syllabus. I couldn't read it, so
I thought the flow was coming from the right.
MR. KELLY: That's hysterical --
CHAIRMAN WALLIS: You're going through
bubbly and slug and annular.
MR. KELLY: But worse, you know it's bad
that it's so highly complicated. But what's worse is
that it's poorly suited for inclusion in the
computational model. I'll briefly tell you what I
mean. Each of these regimes is characterized by a
link and that link is a function of capillary number.
So this is based upon the type of break-up you get if
you take your garden hose out and turn it upside down
and have a jet coming down, when that jet breaks up.
So each of these links is a function of the liquid
velocity, at the quench front. And any of you that
have ever worked at code calculations you know how
noisy that is.
So what that says is the length of each of
these regimes that's been used oscillates with the
liquid with velocity. So, in effect, this type of
scheme amplifies any numerical noise whatsoever. And
in practice, for a forced reflooding case it leads to
very large oscillation that throws most of the liquid
out of the bottle.
MEMBER SCHROCK: I mean this view of
physics ignores the fact that when you look at such
experiments you actually see some large masses of
liquid that get thrown up and then they fall back. So
at any given instant what's happening at some point
above quench front is some liquid going down, some
going up and some hitting each other, of course some
not and net flows. There are aspects of the physics
that are not recognized in this view point.
MR. KELLY: And there are aspects that we
will never capture, even if we implemented second
liquid fuels so you can have some going up and some
down. Because we'll always end up having to treat it
in a time average sense. You know, averaging over some
suitable period which may be on the order of seconds.
But that's --
MEMBER SCHROCK: Well, this is a very
fundamental issue with regard to these equations
altogether. You have variables which are presumably
space and time averaged. No attention given to what
that really has to mean in terms of specific parts of
the two phased domain, where in fact the time scaled
at which you have to be doing the averaging is pretty
long. It's a little bit of a stretch to imagine that
you really have meaningful time smooth variables that
you can work with the same sense that you do, for
example, in single turbine and single phased flow.
MR. KELLY: That's a limitation that, you
know --
CHAIRMAN WALLIS: This is the present
state of the arch you put up there and they're going
to improve it.
MR. KELLY: Well, the first thing I want
to simplify it and come up with the energy and
hopefully, the plan then is to use the theta from the
RBHT facility to come up with a more mechanistic
model. But to go to the one more detail that
Professor Schrock is, that's really out there,
especially in a computational framework where you're
talking of modeling the power output.
And as a result, this shows an example.
This is FLECHT-SEASET 31504 which is the rate
excessive force flooding case. Clad temperature
versus time, this is just above the core mid-plane.
This is the data from three different thermal couples,
and this is the current TRAC calculation. And you
notice this is more than 300 degrees K, and this would
be completely --
MEMBER SCHROCK: Well, it's conservative.
MR. KELLY: It's highly conservative.
MS. UHLE: Extraordinarily.
MR. KELLY: At least for a forced flooding
case and that's because there are in effect these
vapor explosions, if you will, which throw most of the
liquid from the bundle out the top and FLECHT-SEASET,
you know, can point to that as a phase separator so
the water can't come back down. So a one inch per
second case ends up being like a one-tenth of a second
case which with that flow rate we have a very hard
time turning the temperature around.
CHAIRMAN WALLIS: It's pretty good to be
up at 1400 degrees Fahrenheit.
MR. KELLY: So, obviously, we have some
work. This is why we're doing the work, while we try
to apply this to AP1000. And so, obviously,
improvements need to be, we have to reduce the
oscillatory behavior and improve the accuracy of the
prediction. And I'm going to try to do that with
using a simple modeling first, and wherever I can use
bundle data, sometimes tube data to come up with a
simple way of doing this and one that is less
suspectable to oscillation.
I'm now on the last page of my talk and,
hopefully, this is practically finished.
CHAIRMAN WALLIS: Well, you've had your
hour.
MR. KELLY: Yes, I'm afraid so.
And I'm going to talk about incorporation
of experimental results, and hopefully very briefly.
And I've got a little note here, the ACRS
role is I think that would be very good. As we do
these experimental programs and these are our
contractors who are asked to develop models from them,
it would definitely help us to come in front of you,
present those models and get your opinion. And in
effect, for us to have a peer review via you of how
good those models are before they get, you know,
encapsulated in concrete. So this is certainly an
area where I think you could help us. You know, as
kind of as unpaid consultants.
We currently have four experimental
programs. Low pressure, subcooled boiling at UCLA,
phase separation at OSU, which you already know about
since it's been out there, the rod bundle heat
transfer programs at Penn State, and the interfacial
area transport at Purdue and the University of
Wisconsin.
This general program will be finishing in
the middle of the year. The model will be delivered
and they will be implementing the code late 2002.
Phase separation, hopefully that in 2003.
The rod bundle heat transfer, this is one
I want to talk a little bit more about. It's designed
to provide detailed measurements for model
development. It's not simply let's get some more
reflood data, because there's a lot of reflood tests
out here. But there was a lot of thought in how to
try to instrument the bundle and what development
information we need. And that's the example that I'm
going to use on the incorporation of experimental
results.
The reflood tests will be conducted in
mid-2002. There will be 15 of them, roughly 12 or 13
will be for model development. There will be no
constant flooding rate for test cases to look at one
particular regime.
There will also be 2 or 3 variable
flooding rate cases which we'll use for code
validation.
But we're also going to use steam cooling
and drop injection tests, and I'll talk about those in
a little bit. And those will be in late 2002. Then
the data analysis and model development will be in
2003/2004. And at that point we'll have low
mechanistic reflood model in the code to do that.
The interfacial area transport, this
should be viewed as a long term exploratory research
program and the idea is to try to move the level of
the physical models one step closer to something
mechanistic where you're now looking at pebble
coalescence and breakup instead of the static flow
regime model. And so we're due to implement this
model in 2005, however the data is being generated now
we'll be able to use as part of PIRT assessment
program.
And the key thing on this slide is as
these programs end, we hope to start other
experimental programs to take their place so we keep
the level of thermal-hydraulic experiments that we're
funding more or less constant in time instead of
on/off. But replacements to these experiment programs
will come about from code assessment results. We
identify a deficiency in the code in an important
element, can't find the data in the extent database
and get a targeted date, then we'll identify an
experimental program and try to get one started.
So this is the example of how to
incorporate the experimental results. At least the
beginning of that. What I'm going to talk about is
the dispersed flow film boiling agent, which is the
one that we think of as the most important in terms of
the large break LOCA because that's where you turn
around the clad temperature.
In this regime the most important heat
transfer mechanism is forced convection from the rods
into the vapor. To the superheated vapor. But there
are two major unknowns. One is the drop diameter,
which is a rather fundamental quantity and the other
is two-phase convective enhancement.
The drop diameter is primarily important
because of its effect on the vapor superheat. I mean,
after all, that's your sink temperature. You're
transferring heat via conduction through the steam to
a highly superheated steam. So what that temperature
is is very important.
However, it also effects drop breakup on
the grids, the two-phase convective enhancement, as
well as the wall-drop radiation heat transfer.
In reflood the drop formation mechanism is
not known, and every paper you read says something
different. Is it aerodynamic breakup of liquid slugs,
or a breakup of an actual inverted annular column.
Sometimes it may be one, sometimes another.
There could also be wave entrainment
either from waves on an inverted annular core or if
you're in a low flooding rate case where you actually
have annular core below the quench front it can be
waves on that film.
You can also have wall to drop
interactions. A drop can collide with the wall and
bounce off and shatter, or it can collide with the
wall and in effect be blown off by rapid boil, and
that can shatter the drop.
Which of these mechanisms or how these
mechanisms interact to control an average effective
drop size is really unknown.
CHAIRMAN WALLIS: These are all
speculations or fantasy, you mean?
MR. KELLY: Yes.
CHAIRMAN WALLIS: They're not based on
observation?
MR. KELLY: Well, some of them are.
CHAIRMAN WALLIS: They are?
MR. KELLY: Yes. Depending upon which
paper you read, various people say different things.
CHAIRMAN WALLIS: Is that because they've
actually seen it or they've speculated it?
MR. KELLY: Well, some of it is seen. For
example, the annular film and waves on the annular
film comes from a British paper reflood in, I believe
it was a quartz tube.
The breakup of liquid slugs, I don't
really remember.
But, you know, I've been through a lot of
references and you see a lot of different things.
Two-phase convective enhancement, what
this is, you know, we know that the core's conduction
is steam. But if you have a dispersed phase, whether
it happened to be solid particles or drops, that will
effect the heat transfer rate. And now especially in
the case of drops, the act is heat sink, so vapor
sources -- preliminary estimates of data say that this
should enhance your flows convection heat transfer by
20 to 100 percent. But, again, the controlling
phenomena is not known. Is it via turbulent
enhancement?
We know from like, you know, grasped
particles in air if the particles are very small, in
the order of 30 microns or so, they do tend to excite
the turbulence and increase the heat transfer. If
those particles go up to about 100 microns, they damp
the turbulence and decrease the heat transfer.
Our drops tend to be more like 1,000
microns. So how do they interact with the turbulence?
But, of course, there's not one drop size anyway.
There's a spectrum of drop sizes. Some might enhance
the turbulence, some might dampen it. But once you
get up to a millimeter and larger drops, now you've
got drops with significant weight regions which could
generate more turbulence because of that.
Likewise, if you have all these drops
distributed in this hot steam, you change the
temperature profile of the steam.
CHAIRMAN WALLIS: It's dispersed flow
boiling, it's not film boiling. There is no film.
It's dispersed flow boiling.
MR. KELLY: That's true. That's, you
know, just the way it's been. And what we're trying
to say is that the surface is dry.
CHAIRMAN WALLIS: Yes, but what I think
what they mean is the surface is dry.
MR. KELLY: Right. That's what the film
in that context means. If you will, a vapor film.
So those are two of the most important
things or us to look at. And how are we going to do
that with the rod bundle heat transfer facility.
Let's talk about drop diameter first.
And what I've done is basically put up all
of the drop diameter data that I could find in the
open literature, and this drop diameter data from a
reflood test. And, as you know, there's tons of data
for primarily air, water and tube annular mist flow,
but even if you go from one of those papers to the
other, what the correlations for drop diameter are are
different; there are dependencies on physical
properties or even the vapor momentum flux are
different.
So what I've applied are sauter mean
diameter versus pressure reported in a test. ACHILLES
and FLECHT-SEASET are actually bundle reflood tests.
This is the FLECHT-SEASET data all run at about 40
psi. I spread it out in pressure just so you could
see the points, but they're actually all at the same
pressure or almost the same pressure.
These are from a -- high speed group from
several -- about half a dozen different reflood
tests, different flooding rates and so on. It's
actually pretty amazing that the sauter mean diameter
is as constant as it is, just a little above one
millimeter.
CHAIRMAN WALLIS: Six millimeter is a
humongous drop.
MR. KELLY: Yes, that's a problem, too.
And what you have then is water plugging the tube, and
that's why the drop can be carried up. It's the
container wall effect otherwise for those cases the
vapor velocity would be low enough you couldn't carry
the drop up.
CHAIRMAN WALLIS: Even one millimeter
seems pretty big.
MR. KELLY: I agree, especially if you
look at a rod bundle with the grid space, and you go
how can a poor little drop get through.
The ACHILLES tests, those were actually
from two different reflood tests, but the distribution
isn't --
CHAIRMAN WALLIS: So mean diameter, that's
a mean diameter of 6 millimeter. It must mean some of
them are two centimeters. That's crazy.
MR. KELLY: Well, it can't be bigger than
the tube. I agree, those are huge.
These tests, these are rod bundles, these
are tubes. This Hall & Ardron, this was done at CEGB
I think in the early '80s, I don't remember. This was
done at University of California Berkeley by Seban et
al.
This is reflooding --
CHAIRMAN WALLIS: Before they married to
one another.
MR. KELLY: It's hard to know which of
these to believe. But it would appear --
MEMBER SCHROCK: This kind of statement
bothers me, it's hard to know which of these to
believe. These experimentalists are presenting data
from different kinds of experiments and why do you
think that as a code developer you're going to
evaluate which among these that had maybe different
objectives even, is right or wrong? I wouldn't begin
by assuming some are right and some are wrong. I'd
begin by trying to understand why is there this kind
of apparent discrepancy that arises out of these
different kinds of experiments and how does it relate
to the simple or the actual system that I'm trying to
model with this code.
MR. KELLY: No. That's a very good point.
Both of these were tube tests, but one was a quartz
tube, one was a, I don't know if you know the CRE
valve. They were both basically the same kind of
traditions opposed to directed methods of tube
resonance.
MEMBER SCHROCK: If you get into details
of the paper, you'll see that the credibility of the
meaning of a sauter mean diameter for some experiments
may be better than, you know, some other experiments.
MR. KELLY: Depending on the sample size,
that's exactly correct. And what I probably
misstated, mispoke a little -- what I should say is
this isn't solely a function of pressure. And what
you may very well be seeing here, rather than one
being right and one being wrong, there may be at
different values of the vapor momentum flux, and that
may explain the large discrepancy. But from
everything I've seen so far, is the vapor momentum
flux goes up, the drop diameter goes down.
CHAIRMAN WALLIS: I guess the message I
get is that you're looking at all these things, you're
trying to figure out the reasons for discrepancies and
do better at it.
MR. KELLY: Right.
CHAIRMAN WALLIS: At the level we're at
today, we can't get into the details.
MR. KELLY: Right. Most of the current
models that people tend to use in codes are simple
functions of the LaFosse number. So it's a function
of pressure only.
CHAIRMAN WALLIS: LaFosse with gravity in
it?
MR. KELLY: Sigma over G delta rho
squared.
CHAIRMAN WALLIS: Does gravity have
anything to do with the phenomena that's happening
here?
MR. KELLY: Well, what they're saying is
that you can only -- if you're given vapor flow, you
can only get up to a certain size drop, and less, then
use a critical web number of criteria for what that
size will be equated to and you come out with that.
And that's roughly how you come up with that. And
that will give you the maximum size drop, then you
have to make some assumption to get from that to a
sauter mean, typically a factor of 3 or so, but you
know what exactly it is is hard to define. But it
looks like there's a pressure relation here, but that
could be some other reason.
CHAIRMAN WALLIS: If you did the
experiment in a space shuttle, the drops would have
zero diameter is that what you mean?
MR. KELLY: No, because then you would
have to have different non-emitional groups because
you have a different control element.
CHAIRMAN WALLIS: Okay. I understand.
MR. KELLY: And as I recall, you'd tend to
get really large drops.
So anyway, that's the data that's there,
but that is certainly not sufficient to develop to a
correlation level. And the real reason it isn't is
because the data base lacks the information on the
flow conditions and we don't know what the vapor
velocity is. We don't know the vapor density. So we
can't come up --
CHAIRMAN WALLIS: So that's right, this is
an example of why you need your RBHT?
MR. KELLY: Exactly.
CHAIRMAN WALLIS: So can we skip to the
conclusion, do you think?
MR. KELLY: Sure. And I'll just go ahead
and skip the convective enhancement.
CHAIRMAN WALLIS: And the rest, we can
read the summary page.
MR. KELLY: Yes. The real point here is
with RBHT we've tried to design the instrumentation to
give us the information, and Professor Schrock earlier
talked about the mechanisms. One of the things to
look at will be high speed video, and I'm looking
forward to seeing and looking at high speed video over
and over again to try to get a better idea of
physically what's happening.
For the convective enhancement by the
drops, I mentioned earlier there'll be two types of
tests run. The steam cooling test, with steady state
heat transfer forced convection to steam.
CHAIRMAN WALLIS: And you haven't been
skipping. I was asking you to go to the end.
MR. KELLY: Okay. But this is something
unique about the facility.
CHAIRMAN WALLIS: We can't spend a lot of
time on all these different items.
MR. KELLY: Right. And there's the
summary.
CHAIRMAN WALLIS: We'd probably need two
days.
MR. KELLY: Okay. So the code development
associated with the consolidation effort will be
completed in the year 2002, probably by the end of
January.
The developmental assessment will be
conducted throughout calendar year 2002.
We're going to update the interfacial drag
and the reflood models; those will appear in the
consolidated code. The consolidated code will
probably be released at the end of 2002.
And then long term code development and
experimental programs will be driven either by code
deficiencies that arise from the assessment program or
by user needs for new capabilities.
CHAIRMAN WALLIS: So you're counting on a
lot of input from this work -- your subcontract? Do
we need to have presentations from these people during
the year so we can see how they're doing?
MR. KELLY: I think that would be a good
idea.
CHAIRMAN WALLIS: Should we probably do
something like that at sometime in the middle of the
year.
MR. BOEHNERT: Sure.
MR. KELLY: I would prefer coming --
CHAIRMAN WALLIS: Again, you don't want to
us shoot down, let's say, phase separation models just
before you're putting them in the code?
MR. KELLY: Right. And Steve -- Steve's
talks is on the status of these program, but I think
that's a good idea.
CHAIRMAN WALLIS: There's nothing like
speaking to the people who are actually doing the
work.
MR. KELLY: Right. Well, hopefully, we're
going to be --
CHAIRMAN WALLIS: Maybe at the end of the
day, but I don't that we'll have any time. We need to
think about how the ACRS can be more central to you
folks.
Although I think when I look at a schedule
here, I wondered if it wouldn't be better off to --
well, I guess, Steve, you have two presentations.
MR. BAJOREK: I've got to two of them.
The first one --
CHAIRMAN WALLIS: Maybe you should make
the first one, and then we can have lunch and come
back for a second one.
MR. BAJOREK: I think that will do well.
Well, if you pass out the one handout, what I am going
to do is I'll just bring out a couple of overheads?
MR. BOEHNERT: This one first, Steve?
MR. BAJOREK: That one first, please.
MR. BOEHNERT: Yes.
CHAIRMAN WALLIS: You're in the last lap
here, and you've got to make up -- you've got to run
at double speed.
MR. BAJOREK: I'm ready to go now.
But in the earlier presentations, one of
the things you may have noticed that Jack noted that,
in the long run, we're going to be counting on the
code and more to make regulatory decisions. The
accuracy will be much more important to us now then
they had been in the past, because we're going to be
relying on TRAC-M, the developers of TRAC-M to come up
with these decisions as opposed to information that we
had previously been asking from the vendors.
Joel also in his presentation pointed out
in the developmental assessment that's going to be
done over 2002, most of that is being directed at
completing the consolidation showing that TRAC-M can
meet the functional requirements of RELAP, TRAC-B and
TRAC-P. The matrix that Joe put up using primarily
the tests that had been used in the past to try to
develop the code and assess its performance. It's not
necessarily the best set of experiments to use to try
to determine whether we're doing a good job or what
we're weak in, or to really characterize the accuracy.
So what I'm going to talk about now is
assessment and quantification of the performance of
TRAC-M. But in many ways what this really represents
in the elements, I think, another five year plan. The
consolidation effort is going to go on through most of
2002. Through that effort we're not going to be able
to do the total amount of assessment that we would
like to have. So we're looking at work further
downstream, 2003 and beyond. What I'd like to try to
do is layout a better picture of where we think we're
going to be able to go with TRAC-M, apart from the
development of the potential model development that
Joe just talked about.
We see three major elements. One, a
continuing model improvement to get information from
these test programs, having we things out, they are
success oriented. We are assuming that the data that
we're going to get from the rod bundle heat transfer
programs are high quality, and likewise for the phase
separation.
We're talking about those programs that
are all the things that we're going to be doing to try
to ensure that we are going to get the right
information to develop those models.
But I think one of the more important
aspects that we're going to have to address in the
next one or two years is how do we assess the code
accuracy?
We've seen models of the code right now
that clearly don't perform as we'd like them to. We
see TRAC as being "conservative," and RELAP being
nonconservative in the reflood heat transfer.
CHAIRMAN WALLIS: Do you really mean
uncertainty? Is that the same thing as accuracy in
your mind?
MR. BAJOREK: Pretty close, yes.
CHAIRMAN WALLIS: So for the user, the
user needs to come to us to identify some
uncertainties in the use of the code, and accuracy may
be a part of that.
MR. BAJOREK: The way I would break it
down is we're going to be looking at various processes
in the code. We're going to have to assess how
accurate the code --
CHAIRMAN WALLIS: Well, all codes are
probably perfect for one point, if you're on the right
point.
MR. BAJOREK: When we get everything in
there, the issue, of course, is that errors arise.
And one of the things I'm going to point out in one of
the next coming overheads is how we are going to try
to overcome that.
CHAIRMAN WALLIS: So I say, the second
bullet is related to the first. The user uses the
code and there's some uncertainties associated with
that. And that leads to margins and all kinds of
stuff. If the uncertainties are reduced, that could
be model improvement. For certain applications you
don't need any model improvement. But for other
applications you made a lot of model improvement.
It's got to be somehow related to the uncertainties
which are needed for the purpose of regulation.
MR. BAJOREK: That's why as we go through
our model development we'll be relying on separate
effects testing to get this biased uncertainty from
models. But an element that we will build in early on
is how to do those uncertainties propagate a behavior
when you apply them to a PWR or a BWR.
CHAIRMAN WALLIS: So you're going to have
some sort of mathematic or analytical framework for
this? If you know the uncertainties in this current
correlation of the model from separate effects tests,
and then you can predict the uncertainties in the
integral effects tests and so on?
MR. BAJOREK: Yes.
CHAIRMAN WALLIS: And then you can predict
the uncertainties associated with some licensing
calculations.
MR. BAJOREK: Yes. Many times I've seen
in the past we've spent an awful lot of time
developing a model for one process or phenomena only
to find out that when you arranged it in a PWR
calculation, it was only effecting your answer by a
few degrees. I mean, that kind of tells you that your
model development effort is being misdirected, where
as other models --
CHAIRMAN WALLIS: This is risk-informed
code development.
MR. BAJOREK: Risk-informed without
development. But that final piece, seeing how the
uncertainty needed in the light water reactor
application is very valuable, because that allows ut
to redirect our model development. It also allows us
to refine or define new experiments that we need to
do. And that's why what we'd like to try and do with
the assessment element of this is to start looking at
light water reactor applications early on, assess the
uncertainties so that we can account for and correct
those in the model development efforts.
MEMBER SCHROCK: I don't quite grasp the
significance of the separate PIRT-based assessment.
MR. BAJOREK: The difference between what
we are calling a PIRT-based assessment matrix and the
code consolidation matrix is in the overall scope and
how the simulations that were performed in the PIRT-
based give a broader coverage of those processes that
have been highly ranked in the PIRT.
The code consolidation matrix is largely
historical. It picks certain FLECHT tests, some which
are antiquated data, they didn't have the best test
instrumentation in there by way you could assess some
of the code models and correlations. What we would
like to do is to get away from some of these tests
that had been used on more of a historical basis,
broaden that to make use of a broad range of FLECHT
SEASET, but not rely just on FLECHT SEASET, look at
the Skewed test, some of the Cosine tests, the
ACHILLES test and other reflood tests to avoid coming
up with a code where it may work good for FLECHT
SEASET, but not do well for other types of experiments
MEMBER SCHROCK: Well, it sounds kind of
like you have to limit the amount of assessment you
can do, and so here is a way of choosing more
important things to perform the assessment on. But
that increases the likelihood that there maybe
something and it's never been understood these things,
it isn't going to be properly addressed in this new
code version, and it never will be.
MR. BAJOREK: We'd like to try to expand
the matrix so it exercises the code over a much
broader range. In some ways there's also some economy
in doing that.
A lot of the work in developing these
input decks for a certain test facility, in some cases
it takes as much work as it does to set up a PWR or
BWR deck. But when you're only going to be running
one test out of the higher series of tests that can
run in that matrix, you're losing a lot of information
that you may gain by increasing the number of tests
that you look at in that facility. CCTF or SCTF are
an example. The consolidated matrix only looked at,
I think, one or two tests. What we're proposing is to
expand that to look at on the order of 10 or 12 tests
so you examine how well the code can perform as you
change things like your boundary conditions, your
break size, your power distributions both axially and
laterally within the core. See how the code uses --
the sensitivities you can get through the code rather
than just looking at one point.
CHAIRMAN WALLIS: I would like you to move
in the direction of risk based assessment. PIRT is
just some expert sitting down and saying "Gee, you've
got to do a better with more than condensation." I
mean, there's no measure of better job until you come
up with things that you're going to use it for. Use
it for making regulatory decisions.
So PIRT I never felt was a really good
measure of goodness of a code, even if it were used
for that purpose.
MR. BAJOREK: Well, the PIRT's kind of
done beforehand, and it really only gives you some
guidance on what --
CHAIRMAN WALLIS: PIRT is a starting
point.
MR. BAJOREK: Right.
CHAIRMAN WALLIS: But it doesn't give you
a measure of success. And I think you really need to
think more about what is the proper measure of success
for a code.
MR. BAJOREK: Let me jump ahead for that
then.
MS. UHLE: Can I answer Professor
Schrock's question?
MEMBER SCHROCK: I think he said it okay.
That really results in more assessment than less. I
have a feeling that it is maybe limiting the amount of
assessment.
MS. UHLE: I think it's just focusing on
where we're going to start first and then getting
gradually to the lower things.
MEMBER SCHROCK: Yes.
MR. BAJOREK: Paul, if you pass out that
other set --
MR. BOEHNERT: This one?
MR. BAJOREK: Yes, that one. This is the
proposed assessment matrix that will be used following
the code consolidated assessment matrix. If it came
out well in this, we would continue to do tests
looking at tube barometers, types of tests where you
know you can do a hand calculation to come up with the
answer that maybe the code has to deal with to perform
those tasks before it could go on to the others. This
is a way of checking to make sure your latest code
change goes through appropriately. But the difference
between the consolidated matrix and what we would be
doing in what I'm calling this first development
assessment matrix, we would greatly expand what we are
looking at in the FLECHT SEASET facility so that we
could look at how the code performs for a forced
reflood, when we change the reflood rate --
CHAIRMAN WALLIS: What's your measure of
performance? In your two-phase pressure drop here
you've got to do some comparisons. How do you know
when it's good enough? Maybe a factor of 2 or 10 is
good enough two-phase pressure drop. How do you know?
MR. BAJOREK: Part of that comes from what
we get out of ranging the bias and uncertainties at
the light water -- in the light water reactor. So
coming up and let me go -- I'll jump to this, and let
me show you --
CHAIRMAN WALLIS: You really have to do
the CSAU thing and look at how does it effect things
that matter, like peak clad temperature or something.
Then say, have we got a good enough code. Don't you
have to go to the things you're trying to predict for
regulatory purposes and the sensitivity of those are
the things that you look for.
MS. UHLE: That comes out of the fact that
these models will probably be ranked --
CHAIRMAN WALLIS: But the PIRT doesn't do
any of that.
MS. UHLE: Well, sure it does. It tells
you what experts are thinking of.
CHAIRMAN WALLIS: It doesn't tell you
what's good enough.
MS. UHLE: It does in a sense that --
CHAIRMAN WALLIS: What an expert's
thinking is really often self-serving. They say I'm
an expert on flow regimes so you need to do more work
on flow regimes.
MS. UHLE: And then in our first
experiments we focus on those models that people point
out as most important.
MR. BAJOREK: We look at the reflood heat
transfer to determine how well it behaves, and we have
looked at some of the reflood tests and we would see
how those uncertainties behave in the full scale.
Now, if we continue to see very large
uncertainties, that's an indication that we need to go
back --
CHAIRMAN WALLIS: If it effects the
regulatory decision.
MS. UHLE: Right.
CHAIRMAN WALLIS: Yes.
MR. BAJOREK: So if we range the reflood
heat transfer over its broad range of uncertainty
based on how we see it in separate effects, but it
doesn't make any effect anymore on the peak cladding
temperature, that says we should look more at things
like bypass or condensation. I don't think we're at
the point where we can rule any of those out.
MEMBER KRESS: Rather than look for what
range it's asking as measured as how good is good
enough, I think your aim ought to be being able to
capture the uncertainties. And then if you can
capture then, you can say how good your prediction is
with respect to any of the reactions and then your
decision process could factor in those uncertainties
on whether or not it's good enough. So again an
application --
MR. BAJOREK: Once can capture them how
well the code's performing based on the separate
effects then we can see how it behaves.
MEMBER KRESS: So how is the code going to
be able to kick out for you the uncertainties.
MR. BAJOREK: If we don't get to that step
and we see a large change or no change in the light
water reactor, you don't know whether it's because the
code is doing input or not or whether it's exhibiting
the right sensitivities.
MS. UHLE: Another thing, too, focusing on
the separate effects test is the fact that if you just
focus on the integral effects test you're not sure if
the answer isn't changing because of compensating
errors. And that's what the separate effects tests
really highlights.
MR. LAUBEN: No, the point is that there
is nothing like the regular development of --
CHAIRMAN WALLIS: No, I'm saying that I
think the PIRT is based on experts, the wrong experts.
MR. LAUBEN: Right.
CHAIRMAN WALLIS: They're not your
customers. They're just the people who are looking
for work. They're the wrong group.
MS. UHLE: Any PIRT contributors here?
All right. So any PIRT you were involved in we'll
throw out.
MR. LAUBEN: But if you were to go through
the process we talked about today, you'd start out
with some kind of PIRT and during the process you'd
focus in on, at the highest level, like you were
saying the ability to predict a regulatory effect on
the peak cladding temperature, and all of the top
level things. The PIRT may change. The PIRT may
start out as something, and what is critical changes
throughout your whole process.
MR. BAJOREK: And it does. I mean, if you
look at the PIRTs that are designed for conventional
PWR, versus AP600 or AP1000; there are small but
perceptible changes in all those, and what's important
in one transient versus the next --
Our problem is making sure that the code
can deal with those things which people have deemed to
be very important and then can also deal with those
things which are deltas between plants that have been
looked at in the past.
Now, I think part of the problem in this
assessment, I think has just been pointed out, is a
lot of folks have focused on solely the peak cladding
temperature as being your sole measure of a code
performance. And what I did is I grabbed a couple
rolls of technical papers and, actually, I took one
out of CSAU NUREG for example, how does your code
behave. And the common way of doing it is looking at
the peak cladding temperature from the scout point.
A plot, if you were, where you predict the PCT as
higher than the measured, you deem that it's being
conservative and say that your code's conservative
forgetting the fact that there may be other things
going on in these experiments, CCTF and SCTF in steam
binding and the steam generators that may be
contributing to the performance of your core heat
transfer.
Another way would be taking these tests,
mix them in with separate effects tests, which is done
over on this figure on the left hand side, and use
that to get a gauge of your code performance, or in
this case as this had been designed to, is well let's
just get a delta PCT and you would simply put that on
as an adder towards some calculation that you would do
for, in this case, the PWR.
I think the perception now, and correctly
so, is that approach is incorrect because it doesn't
deal with compensating errors. It doesn't deal with
new ranges and conditions and tells you nothing about
whether you're getting things like super heat, drop
break-up correct, all of the intricacies of reflood
heat transfer that go into calculating that peak
cladding temperature.
Now, we intend to expand the test matrix
that we're going to for the separate effects tests,
but at the same time which when you get away from this
type of a measure of the code performance. This is an
example of typical practice, and the one I just showed
you -- I've got my greater than and equals sign in the
wrong direction here. The one where I just pointed
out is to take a look at solely the peak cladding
temperature, it says your coding is conservative if
your predicted is greater than the measured.
Now another way, and Joe alluded to it in
his presentation, is to look at the one model in
reflood that's perceived as having the greatest
effect. Okay. This has been done by taking a look at
the dispersed flow film boiling heat transfer
coefficient; defining a bias and an uncertainty. And
in this particular application then the uncertainty in
that particular model was used to range at full scale
in order to get delta PCTs in the full scale case.
PWR in this case.
We're going to take advantage of more
detailed test data like we're getting out of the rod
bundle heat transfer program and information that we
can glean from other test programs to increase the
total number of peak performance parameters before we
can claim success in any one of the models, and I'm
going to reflood as an example.
Our approach now is we can use multiple
parameters is to try to characterize specific models
within the package and the package in total. Okay. We
will not rely on simply peak cladding temperature as
the sole performance indicator. For reflood heat
transfer the type of things that we would get out of
the assessment after we have done the simulations and
comparison to data, the FLECHT SEASET, that larger
number of tests; the FLECHT Skewed, the FLECHT
ACHILLES, the other ones that I have listed on there,
is to look at break things up into heat transfer
regimes. Look at those periods where the test and the
code were predicting steam cooling heat transfer, and
use this as a performance measure by defining a bias
and uncertainty essentially to characterize how well
the code is characterizing and calculating in a single
phase performance. We would still do the dispersal of
film boiler heat transfer coefficient as we've done in
the lab, in case you're not aware of.
Joe noted that near the quench front,
okay, we also have some very important precursory
cooling. And we want to know whether the bias and
uncertainty in the models that we develop and put into
the TRAC-M are reasonable compared to the experimental
data that we get out of the rod bundle and we can also
get out of some of the other tests. And this is one
case where we might want to jump very quickly to take
this biased uncertainty and use those or study those
in a light water reactor application and give us an
indication should we be looking more closely at
inverted annular flow, okay? Or, should we continue
to focus on steam cooling dispersed flow film boiling
which has been more typical of the past.
The answer to that in those simulations
would be whether we're seeing very large uncertainties
in the light water reactor application, very large
delta PCTs. That would be an indication that the bias
and uncertainty that you are imposing on the code by
a model selection and model development would be
unacceptable. It might mean another experimental
program or it could at least mean you would have to go
back, sharpen our pencils and come up with a better
model and do some additional assessment.
MEMBER KRESS: Is the plan incorporating
these biases of uncertainties into the code itself and
combining in some way with the Monte Carlo, for
example?
MR. BAJOREK: In the long run, yes. Right
now we don't have any plans to put in into the input
structure in the TRAC-M the way of incorporating these
biases or uncertainties easily.
MEMBER KRESS: It seems to me like that
should be your eventual goal?
MR. BAJOREK: I think if we start to see
it -- we did want to have some kind of input or some
way within the code structure that we could range
these things easily without depending on either the
developer or the user to actually go into the code and
hard wire the changes, which is the way I've seen this
thing done in the past.
MS. UHLE: A good thing about the
modernization, too is the architecture. The physical
models are isolated from any of the -- associated with
the alphanumerics. And so the correlation in a
specific sub routine is either divided or multiplied
by that value, and have that propagate through the
answers.
MR. BAJOREK: It has also quite helped us
to get away from relying on that group of experts that
helped develop PIRT. Because once we try to develop
a larger range of performance perimeters, and have
really to range those in the light water applications,
now we can go back and say ah-ha, this should have
been on your PIRT and this was missed or hopefully you
guys did a job.
MEMBER KRESS: I worry about arranging
them individually one at a time. That's why I
mentioned in Monte Carlo, you can get away from that.
MR. BAJOREK: The way we had done that at
Westinghouse was to look at things one at a time and
then develop a response service methodology to try to
incorporate how combinations of things can change.
That also was driven by a couple of different things.
One, it was always nice to go to the user
and say "This is what you're going to do because this
is what was approved," very clear cut.
Another approach, and I think that has
been used more in Europe and we are going to be
looking at that in the long run, is I think is a GRS
sampling approach or refer to it as a German sampling
approach. I thought that was Oktoberfest.
But what this does is it looks at a broad
range of uncertainties and simultaneously picks and
ranges multiple perimeters, and puts that in your
simulation, samples that distribution many times that
gives you an uncertainty in your peak cladding
temperature, your equivalent clad reaction and also a
confidence interval. If you don't like that
confidence interval, do it more times.
Now, the nice thing is that it seems as we
spend more time in these meetings computers continue
to get faster. And what would seem, you know,
absolutely insane ten years ago, making a hundred PWR
or BWR calculation, is now something that can be done
in a reasonable amount of time. So that type of
approach now I think is something that can and should
be looked at in the long run.
But anyway, we're going to break up, for
example, reflood into multiple performance perimeters,
in some we are going to look at specific model and
processes. I wanted to add a couple on here to try to
address the hydraulics, although it doesn't
individually get at flow regime transition or
interfacial drag, but carry over fractions. Rather
than just taking a look at mass affluent and what the
code is predicting, it applies an uncertainty for as
many of these tests that you can so we can determine
if the code is doing a good job or not in calculating
things like entrainment drop size.
Level swell, or another way I would say it
interfacial drag below the quench front. And see for
a given amount of mass has the quench front propagated
too high into the bundle.
And in those characterized individual
models or call them packages, we can use things like
quench time, clad temp and steam temperatures, these
are in the program that was developed at Penn State
called ACAP that essentially goes through and takes a
look at a predicted trace versus a measured trace and
gives you statistics on how well that curve
corresponds to one another.
I'd like to think it a little bit more as
the integral of this curve behaving much like the
integral of the other curve. This, again, starts to
get closer towards the peak cladding temperatures
you're looking at things that's an aggregate, but we
think by defining several key performance perimeters
and making our holy grail the idea that we're going to
get all of these simultaneously in some reasonable
bias and uncertainty where reasonable at this point
still is yet to be defined, because when we go through
the first cuts and range those in PWRs and BWRs,
that's going to start to tell us what is reasonable,
whether we're looking at hundreds of degree change in
clad temperature or a few degrees.
Most of this work will not begin until
late 2002 with the release of the Rev 0.0 version.
2002 is going to be primarily those tests that are
being used in the consolidated assessment. What I
wanted to note on here is what we would be doing is
expanding the database both in the total number of
tests that would be looked at in an individual
facility and in the total number of facilities that
would be factored into the assessment.
Some of these tests will be done in 2002
once we get close enough with the Rev 0 version,
because one of our first applications is going to be
the AP1000 large break LOCA. So we not only have the
work at performance for reflood heat transfer, but
also things like bypass, we're going to be very
interested in the performance of the code for how well
it does for direct vessel injection. So we would be
look at tests like UPTF 6 and 21 phase D to get the
direct vessel injection, and also one of the CCTF
tests that also gets --
CHAIRMAN WALLIS: All right. Can you
finish by 12:30?
MR. BAJOREK: Yes.
Integral tests also captured in the
assessment matrix would expand the number, the total
number of facilities. What we would be looking for
there is, does the code give us the type of
sensitivities that were observed in these various
tests. What would we do if we would look at, for
example, SCTF and the difference between flat radial
shape and the very peak radial shape, in the sense
that we are able to get that same type of a variation
in TRAC-M. We wouldn't get that if we just looked at
one test.
And you can see some of the other
sensitivities we would hope to get out of the integral
effects tests.
The eventual goal then is defining the
uncertainty for a large number of models, develop the
capability of range and base and assessing their
importance in the full scale plants, peak cladding
temperature and their effect on normal clad reaction.
As we start to see plants being operated, they're now
staying at higher temperatures for longer periods of
time. Our concern from a risk based regulation is
that maybe peak declading temperatures is what we're
going to have to look at in the future, so we're going
to have to start looking at clad ductility and clad
reaction rate in a lot more detail than it had been in
the past.
As we mentioned, if we start seeing large
uncertainties in the light water reactors, that's an
indication that we either have to look at test data
closer or we have to go back to develop better models
for the process we're interested in.
We see this as being one of our major
activities over the next 3, 4 or 5 years.
And by way of summary, we're going to
expand the consolidated test matrix to look at a
larger number of conditions, a larger number of test
facilities. We're going to use this quantified code
to model accurately and engage what goes on in the
other plants.
CHAIRMAN WALLIS: You have a lot on your
plate.
MR. BAJOREK: There's a lot there, yes.
One of the things we are going to try to
do in 2002 is automate the process. It's a lot of
work and there's a lot of comparisons to data. If we
do a good job on the first few tests, capturing the
scripts to do the comparisons, setting up the methods
to run these things in mass and do comparisons to
mass, we may save -- we'll definitely save ourselves
a lot of grief and agony further downstream.
MEMBER FORD: You've got the data
scattered around the one to one correlation like that
what is your matrix of success?
MR. BAJOREK: It's going to be in the
several parameters that were defined for reflood
rather than PCT. The matrix would be to have all those
at a reasonably small bias and uncertainty. Now,
reasonable, I think in the past people have basically
looked at 5 or 10 percent in bias and uncertainly on
the order of 30 to 50 percent. A lot of that just has
to do with the scatter of the experimental data.
MEMBER FORD: I was about to say that
surely that the scatter is obvious in the experimental
data.
MR. BAJOREK: Right.
MEMBER FORD: But your model should be
able to predict that step.
MR. BAJOREK: For the different condition,
yes.
MEMBER FORD: Well, for the -- that
scatter is presumably due to uncontrolled experiments,
but you can quantify that, the degree of lack of
controls. So can you not -- would you not -- your
matrix of success be that you can bound your observed
scatter? Not only in the experiments, but also in the
reactor? I mean that's the uncertainty --
MR. BAJOREK: You would hope that if you
define, let's say, bias and uncertainty in a model,
when you apply that in the separate effects
simulation, you also can show that you've bound or you
-- excuse me. You bound it in the separate effects
tests and you are confident finding delta C in the
whole scheme.
MEMBER FORD: The wider reasoning behind
my question is that asking for the matrix of success
if a licensee comes in with their own code, do you use
your code? I mean, I know why you're touting your
code, to be an informed reviewer, but at what point do
you say this model is no good based on a matrix like
that in comparison to the observational query. Does
yours do better than he or --
MR. BAJOREK: That's what we're hoping.
MEMBER FORD: And if that happens, then do
you say he can't use his code?
MR. BAJOREK: No, because I think what
happens is if you do a good job on your code, you
should have a relatively small uncertainty when it's
propagated. If you did a poor job on the code, that
should grow.
So, if you come in with a code that does
not perform well against separate integral effects
tests, the price you pay is a larger uncertainty of
the whole scale application. My twist on that is if
your code doesn't have the right sensitivity, I guess
that's a question we have to look into.
CHAIRMAN WALLIS: Steve, we're going to
see you after lunch?
MR. BAJOREK: Yes.
CHAIRMAN WALLIS: What I propose is that
we break for lunch and we get back here by 1:00? Can
you do that, have a quick lunch.
MR. BAJOREK: How much time do we have
after lunch?
CHAIRMAN WALLIS: 2:30. We'll be back
here at 1:00.
(Whereupon, at 12:33 p.m. the meeting was
adjourned, to reconvene this afternoon.)
A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N
(1:08 p.m.)
MR. BAJOREK: This afternoon what we'd
like to start doing then is looking at and reviewing
some of the work that has been done over the past year
on the experimental programs that we're relying on
right now to solve some of the major thermal-hydraulic
issues, and also to give us some additional data for
the code development.
The ones that we're going to talk about,
a couple of these we may move quickly because we've
talked about these back in July, are:
APEX, work that has been going on there to
address the pressurized thermal shock; work that has
been going on at the PUMA facility, Purdue University.
The work that has been going on to take a
look at critical flow, and we anticipate using the
facility to take a look at the BWR boiling
instability, the flow instability.
The rod bundle heat transfer program at
Penn State.
ATLATS or the phase separation work that's
being done also at Oregon State University.
I'm going to present some work that has
been recently given to us by Vijay Dhir at UCLA
looking at subcooled boiling.
And finally, we'll wrap up taking a look
at the interfacial transport project that's being done
at Purdue by Dr. Ishii, also Kajasoy at the University
of Wisconsin, Madison.
But for the work that's been done in 2001
in APEX, APEX in late 2000, maybe a little bit
earlier, had been modified to look much like a
combustion engineering unit. It took advantage of the
fact that the APEX facility in its original format for
the AP600 had a 2 x 4 loop, the pumps were replaced.
Excuse me. The can pumps were replaced in the APEX
facility with loop seals and pumps so that it would
look much like Palisades and Calvert Cliffs.
Most of the experimental work that has
been going on in APEX over 2001 has been designed to
take a look at PTS issues.
Now, we presented a lot of this
information in July of this year when we also got to
see a test at APEX. And I've got a couple of
overheads to summarize the PTS work.
Most of the work that is going to be
planned at APEX for 2002 is going to be directed
towards the AP1000. Dr. Rais was recipient of a DOE
MURE grant earlier this year. This gives him funding
now to modify the APEX facility to replace the heater
core, enlarge the pressurizer, change the core makeup
tanks, add some additional instrumentation so that it
looks much more like the AP1000. And that's the work
that will be going on later in 2002.
Now the PTS work that was being done at
Oregon State was the central part of three overall
components to take a look at PTS. OSU, the APEX
facilities, was used to develop the experimental
database, look at downcomer mixing effects. This was
accompanied with RELAP and REMIX calculations to try
to gauge how quickly these plumes would dissipate in
the downcomer. This was accompanied by a thermal-
hydraulic uncertainty evaluation that was done at the
University of Maryland.
CHAIRMAN WALLIS: Is that still being
done?
MR. BAJOREK: It's finishing up right now.
Following the meeting in July/August time
frame, Dave Bissette decided that they needed some
additional tests to add to the ones that had been
previously done. They started doing those in
September/October time frame. As of October they were
almost done. I think they still had a couple more to
do.
CHAIRMAN WALLIS: Final report is in
December?
MR. BAJOREK: December. Yes, the end of
this year.
The work in the facility as it is right
now, scaled as I mentioned to the CE plants, the work
that had been done --
CHAIRMAN WALLIS: The question is whether
or not this has been adequate to resolve the PTS --
they've done something to my mike? They took it away.
Someone took it away. Oh. Yes.
MR. BAJOREK: I believe it is. Is Dave
Bissette here? I think he's left. But --
CHAIRMAN WALLIS: There's a mike here.
MR. BOEHNERT: No, it's the table mike.
CHAIRMAN WALLIS: There's a mike here. So
we're okay?
MR. BOEHNERT: Yes.
MR. BAJOREK: From my understanding,
they're going to be able to wrap up the tests this
year, issue the final report and I believe that is
going to resolve the PTS issue, which leaves us for
upcoming events.
The early part of the year will be
occupied primarily with finishing the testing, writing
the final report. Excuse me. Not the end of this
year, that's going to be due the end of January. But
starting the end of this year and into most of 2002
leading towards the end of the summer, the facility is
going to be modified. The larger-diameter heater rods
in the core are going to be replaced with smaller-
diameter rods. They're going to put in a new data
acquisition system.
The pressurizer in the AP1000 is
substantially larger than it is in the AP600. That's
being replaced. Likewise, the CMTs. CMTs are larger
in the AP1000, they also have a different type of
orifice to reduce the form loss from the CMT into the
DVI line.
CHAIRMAN WALLIS: Do you know what the
licensing schedule is likely to be for 1000?
MR. BAJOREK: Right now we're scheduled to
issue an SER early next year. I'm not sure exactly --
CHAIRMAN WALLIS: So these tests will come
after the SER has been issued?
MR. BAJOREK: I'm sorry. SER for phase 2
of the review. Phase 2 of the review is taking a look
at the codes, for their adequacy, taking a look at the
test and the analysis program. And we're going to be
issuing our opinion on those, probably March or so of
2002.
Phase 3, Westinghouse decides to go ahead,
they would be issuing their analysis, the finalized
design and then going through the rest of the review.
I think the SER for that would have to be sometime
late in 2003/2004.
CHAIRMAN WALLIS: So your results will be
timely enough input?
MR. BAJOREK: Yes. Yes. It seems very
aggressive and ambitious, but they're hoping to do all
of this modification to the facility and be able to
begin hot down testing the end of next August. If that
were the case, testing would begin later in 2002 and
probably continue well on into 2003.
CHAIRMAN WALLIS: Now is the government
doing some analytical work to figure out what key
tests need to be run?
MR. BAJOREK: DOE asked us several months
ago to comment on the text matrix. They made it clear
it is their test. We gave them some recommendations
based on previous tests that have been run in the
AP600. The ones there that had been the most
interesting from a licensing standpoint were the DVI
line breaks, cold leg breaks, okay, where you had
multiple failures and failures of the ADS-4 system.
Those are the ones that gave the minimum inventory in
those tests. And we would presume that those tests
would also generate the minimum inventory in the
AP1000.
At the top of the list is the DVI line
break. That one, by far and away, seemed to be
generate the minimum inventory.
One of the things that I've been involved
with over the summer and the last couple of months has
been in a scaling analysis for the AP1000 in the test
program. Part of our concerns stemmed with what will
go on in the facility -- or more, the full scale plant
during this ADS-4 period.
The DVI line break is clearly going to
make entrainment in the upper plenum pool, entrainment
in the hot leg, into the branch line much more severe
than it was in the AP600. Going from AP600 to AP1000,
that's a 73 percent increase in the core thermal
power.
The vessel is the same diameter. The hot
leg is the same diameter. So having this additional
core power is going to greatly increase the
superficial velocities during the ADS-4 period and
also during the long-term cooling period. So we're
looking at that.
We did make a recommendation that they add
instrumentation to try to get the branch line quality-
-
CHAIRMAN WALLIS: That's the sort of thing
I had in mind. If you'd thought about what are going
to be the big differences that we need to worry about,
therefore design the experiment so they focus on the
right thing?
MR. BAJOREK: Yes. But those tests will
start later in the year and I look forward to seeing
some of those results.
PUMA is the integral test facility that
represents the SBWR. It's located at Purdue
University. It's an integral test facility that has a
reactor pressure vessel, internal components to
represent the core, downcomer, chimney and separator.
Most of what has been done over 2001 has
been used using the the facility as a separate effects
test. Now, this also stems from work that was noted
in AP600, again during this ADS-4 blowdown period.
Rather than critical flow at higher pressures being
the most important break flow phenomena or range of
conditions, during the ADS-4 we have critical flow at
a relatively low pressure.
During the AP600 analysis using RELAP it
was noted that one of the deficiencies in the code was
the performance of its critical flow model during this
lower pressure period. They've been gaining some
additional information in the facility corresponding
to these lower pressures, making use of some advanced
instrumentation.
The long range intended use in 2002 is to
start to look at the BWR flow instability problem.
CHAIRMAN WALLIS: You're saying that PUMA
is going to be used to look at critical flow at low
pressures?
MR. BAJOREK: It has been. It has been.
CHAIRMAN WALLIS: It has been?
MR. BAJOREK: And what I'd like to do now
is Weidong Wang has a few overheads to describe that
work.
MR. BOEHNERT: What's the issue with flow
instability? Has it been looked at or -- are you
going to talk about that?
MR. WANG: Yes.
CHAIRMAN WALLIS: I suppose that these
valves are going to be tested anyway for ADS-4?
MR. BAJOREK: Not the ADS-4. They've
tested the ADS 1 through 3.
CHAIRMAN WALLIS: Oh, it's too big to
test?
MR. BAJOREK: Right.
CHAIRMAN WALLIS: Blowdown from a valve
isn't something you predicted, right?
MR. BAJOREK: Yes. The AP1000 you get a
preset line size --
CHAIRMAN WALLIS: And it's like a straight
pipe?
MR. BAJOREK: Yes.
MR. WANG: My name is Weidong Wong. I'd
just like to give you a little overview about PUMA
project.
I will basically deliver an overview of a
PUMA project and also talk about critical flow and why
we do that, and also inflow instability and the status
for the plant for the coming year.
This PUMA facility is the only operational
facility for the next generation SBWR in the United
States. And the facility is a scientifically scaled
from SBWR and it has extensive instrumentations, over
500, for flow void fraction. And Steve just went over
all this. And I just give you a few pictures,
cartoons that let you have a better idea about what it
looks like.
This is a schematic of what this PUMA
facility. And they have a reactor vessel just very
tall like a pen here, a pen. And this is containment,
and it's a compression pool, basically it's a separate
component and connected by pipes.
Just give you a few pictures so that you
know what we are talking about.
And this is the size of the dry well
containment. You do not have these pictures because
I have difficulties because it's white -- black and
white. And it's not real clear. And here inside is
the vessel and the people are working here.
So you will see it's a pretty large
facility.
And this is a control room, and there are
people working there. And they have extensive
instrumentations and they're all monitored by these
computers or televisions, because we can see the
bundle and the boiler fraction goes through core
vessel or in the compression pool.
And Y which is for the critical flow, as
Steve just mentioned about, actually we know critical
flow from the light water reactor is important under
low pressure, because either that they are all AP600,
they have automatic deprivation systems.
And at a low pressure, basically
mechanical non-equilibrium for the liquid phase and
the vapor phase, so the last thing is that can be
large due to density ratio. And also thermal non-
equilibrium can be large.
And for our code TRAC-M and RELAP5 we note
basically it's assessed for the pressure above 500
psi.
And the shortcomings for the previous
tests, first of all, we know it's not -- they do not
have a detailed in-line measurement for the critical
flow and also no systematic experiment to address the
mechanical non-equilibrium and thermal non-equilibrium
and the pressure effects.
And I tried to quickly go over some
examples results and then give you the conclusion,
because of time.
This is one of the example results.
Pressure effect for the slip ratio. You will see from
here the quality -- okay, this here different
pressures. With the low pressure like for 30 psi with
the experiment we have to go either from 30 psi to 150
psi. And you'll see this slip ratio can be very
significant here.
CHAIRMAN WALLIS: This is for what? Flow
through a nozzle or something?
MR. WANG: I have two plugs, actually. We
plug both for nozzle and for the orifice. For this
particular one, it is for the nozzle. And basically
for the nozzle and for the orifice we saw the same
trend.
And this can tell us, you know, for the
the AP600 application -- first, we had difficulties
with original critical flow model. Then later we
developed some temporary or interim critical flow
model. Use Henry Fausky, which is a homogenous type
of flow model. And then here you will see, at a very
low pressure, the slip ratio can be high.
MEMBER SCHROCK: Of course, your earlier
experiment wouldn't be very good for flashing critical
flow, would it?
MR. WANG: We have went through --
basically we tried to study this mechanical non-
equilibrium, thermal non-equilibrium, and for chemical
non-equilibrium we used air and water. And for the
thermal non-equilibrium we used super-cooled water and
the flow between we basically used saturated water and
steam to make this experiment.
MEMBER SCHROCK: You misunderstand the
intent of my question. As liquid is vaporizing it's
adding more momentum mass to the vapor phase. That
phenomenon doesn't occur to your water experiments.
So the question, is the adequacy of information from
air-water experiments in flashing steam flow
experiments.
MR. WANG: Well, certainly, I'm not sure
for the answer, but I think that we have a
parametrical study basically for this -- our objective
to get mechanical non-equilibrium and here, if we have
a pressure, it's high enough.
CHAIRMAN WALLIS: Well, this slip ratio is
very dependent on the flow regime. If you have
flashing mixture which is breaking up into droplets
because of flashing, it's very different from
something like an annular. So you have to be pretty
careful about duplication.
MR. WANG: This slip ratio actually is
measured above the break point. It's not really at
the choking plane And basically we've --
MEMBER SCHROCK: Measured where?
MR. WANG: Above the break point. This
measured in -- we measured the void fraction and then
we measured the quality. And these void fraction, we
measured it by impedance meter, and this quality is
computed by the inlet of this critical flow. And from
this correlation we calculate -- from this equation we
calculated the separation and tried to see under this
low pressure condition this slip ratio of water
relation or with slip ratio with this -- with
pressure. And certainly this one can not really
represented as the choking plane but it can tell you
something about the slip ratio. It's important at
this choking plane.
I'll give you another example for the test
results. Subcooled water. And from here I have
showed 150 psi for the orifice and a nozzle. And
therefore we focus on one of this same pressure and we
noticed that for the nozzle have a higher critical
flow mass rate. And for the orifice -- for the
orifice it is smaller.
CHAIRMAN WALLIS: What area is this based
on? This is based on the total area of the orifice
hole?
MR. WANG: Yes.
CHAIRMAN WALLIS: No vena contracta or
something?
MR. WANG: Right, for the orifice, right.
CHAIRMAN WALLIS: You expect something
like this from the contraction.
MR. WANG: Right. And here, we tried to
-- basically tried to see what is important factors
for this thermal or mechanical non-equilibrium and we
concluded basically -- I say we, this project is done
at Purdue University -- and we have concluded because
say if you only have this orifice and the orifice,
since it's short, and it doesn't have much time for
the liquid to evaporate. And we expect some kind of
higher critical mass flow rate, but we see here it's
smaller. And we conclude that basically it's a
mechanical non-equilibrium is more important than
thermal non-equilibrium. This is all that we wanted
to say here.
CHAIRMAN WALLIS: When you compare a
nozzle with an orifice, how do you decide what's the
effective flow area for the --
MR. WANG: We use the same flow area.
CHAIRMAN WALLIS: You don't have a
contraction coefficient?
MR. WANG: Yes, we do not. We'll use
that, but we expect, of course, for this orifice you
have a higher loss, but here we focused on the
orifice, first for the geometry basically for the --
for the nozzle, the lighter liquid -- lighter vapor
have a high acceleration. This is also is not a
explanation. But we really try to here to see what
the effect of the thermal or mechanical non-
equilibrium
MEMBER SCHROCK: I don't think you've got
it right. Critical flow rate is going to depend so
strongly on where the flashing begins. The orifice is
not going to behave like a nozzle. I think you're
seeking an answer to a question which is -- may be a
reasonable question, but I don't think your approach
is going to get you there.
MR. WANG: Okay. We'll feed it back this
to Professor Ishii and we'll have more discussion and
try to get back to you.
And these tests show you the examples
related to this, because the code cannot really
predict the data well. This is the RELAP5 prediction
for this critical flood, 30 psi for the orifice case.
And also we should check on here there's some bigger
problems.
Just gave a summary of this program.
And for the critical flow we have
basically 15 to 25 percent higher flow rate for the
nozzle than orifice. And we notice a larger slip
ratio with the lower upper-stream pressures.
MEMBER SCHROCK: In the Purdue work is
their critical flow measurement using water?
MR. WANG: Basically we have -- the upper
stream is controlled and we use a steady state.
MEMBER SCHROCK: Well, you've shown us
data for air water systems. And I would argue, just
in general, one should not expect to get critical flow
phenomena where splashing water is based on air/water
measurement. They're quite different systems.
CHAIRMAN WALLIS: Well, I think that the
subcooled water test must be for water alone.
MR. WANG: Subcooled water, yes it's for
water alone.
MEMBER SCHROCK: Water alone?
CHAIRMAN WALLIS: So what you're telling
us is that the code's probably not going to give a
good prediction for this ADS-4?
MR. WANG: Right. And we tried to get
some data and see, in the future model development, we
use this data to develop some better model or at least
we have some data here and maybe for the critical flow
models there's some adjustment we can use to improve
the prediction.
CHAIRMAN WALLIS: Presumably, AP600 was
licensed on the basis that the critical flow
predictions were okay.
MR. WANG: We do have done some work to
predict this -- improve the AP600 and --
CHAIRMAN WALLIS: After they built it,
then they have to open ADS-4 on a running reactor?
MR. STAUDENMEIER: Yes.
CHAIRMAN WALLIS: I thought they were
going to be full scale tests.
MR. STAUDENMEIER: For the valve
CHAIRMAN WALLIS: That's going to be the
proof of the pudding.
MR. WANG: Thank you, Joe. And also
actually, maybe the flow regime use is --
CHAIRMAN WALLIS: I'm suggesting that
since this is critical flow, maybe Professor Schrock
should see whatever reports are coming out of Purdue.
Can you do that?
MR. WANG: Sure. Right now I have a draft
report, so I will try to -- but right now we only have
a draft report, so it's not --
CHAIRMAN WALLIS: So it's a useful time to
review it.
MR. WANG: Right. Actually, I try to get
it to you as soon as possible.
MS. UHLE: We'll give that to you.
MR. WANG: And the flow instability, it's
planned for this year and next year. Why we do that
is because we saw some flow instabilities for the
operating PWRs and also for the AP600. Small break
LOCA we noticed there's a lot of flow instability.
And we would like to get some data to assess TRAC-M.
For advanced BWR based on natural
circulation, for example, like SBWR. And this --
based on natural circulation pressure is more prone to
instability, especially during start up because
there's no forced flow which you have a -- if you have
forced flow, you reduce a chance to have instability.
And flow is determined by natural circulation and void
fraction. And the power affected also by the
fraction, as I've said, some feedback and there's some
strong covering.
And for the largest scale experiment which
takes data from simulated material is not available
and the effect of void fraction, feedback and also
time lag for this -- convection time lag is not
studied -- that is our objective to try to --
CHAIRMAN WALLIS: PUMA isn't a nuclear
facility, is it?
MR. WANG: It is not, but we try --
CHAIRMAN WALLIS: How do you do void
reactivity power --
MR. WANG: We try to use some parametrical
studies. For example, we can measure inside the core,
we can measure the void fraction and from the void
fraction, we can use our kinetics code to get some --
the power feedback on the zed power.
CHAIRMAN WALLIS: Can you program the
power to reflect the void?
MR. WANG: Yes. And for time lag also we
do similar trick. Say, for this use electrical rod.
It's not nuclear power and the time lag, they will be
different, but we will try to find out from the
nuclear fuel and for the time lag how much, then we
will try to control the electrical power to delay the
--
CHAIRMAN WALLIS: So electrical power was
a shorter time response --
MR. WANG: Right, it's much shorter.
CHAIRMAN WALLIS: -- so you could program
in?
MR. WANG: Try to delay some certain
amount so that it could match the feedback.
And this is basically our objective to try
to obtain some instability data from this larger
facility at the low pressure and also obtains
experiment data for BWR and low point when reactivity
you have a feedback. And evaluate TRAC-M for the
ability to predict three different types of
instabilities like density wave, flow excursion and
the flashing-induced instabilities.
And also try to see the accuracy of
prediction based on stability boundary and amplitude
and frequency. And the ability to model effects of
neutronics and thermal conduction time lag on
instability.
And where we thought -- just a summary.
MEMBER LEITCH: How do you get low
instability with power feedback. I thought it was the
power feedback that basically led the PWR into an
unstable situation?
MR. WANG: Basically in say we cover
constant power. And if it's a perfect steady state
it's fine. But if you have some perturbation for the
inlet velocity, then instability can occur with
certain geometries. For example, like density wave,
if you have some inlet velocity perturbation, then the
boiling lengths will be changed. If the boiling
lengths has changed, then there's a pressure drop
across the channel because a two-phase flow in a
single phase pressure drop it's kind of automated and
it contains. And that effect can propagate into the
system if system just have like out of phase and you
have density wave --
CHAIRMAN WALLIS: I think if you change
the words a bit, if you said data on BWR transient
flow response, but you could sort of vary something
and then look at the transient response. It doesn't
have to be unstable just strictly to produce a
transient response which could then lead to
instability when coupled. It doesn't have to be
unstable for you to measure these kind of times in
response.
MR. WANG: And basically I just revealed--
actually I have a summary --
CHAIRMAN WALLIS: Does this feed into a
summary of regulatory response to say BWR, how it
operates?
MS. UHLE: Yes.
MR. BAJOREK: It's part of the synergy
program.
CHAIRMAN WALLIS: Part of the synergy
program. We'll find out two years later whether or
not we made the right decision on it.
MR. BAJOREK: According to General
Electric they stay away from those regimes where they
would get these instabilities.
CHAIRMAN WALLIS: According to General
Electric there's no problem at all.
MR. WANG: Well, in summary, for the
inflow instability -- actually we only can say the
status -- right now we have done some analytical study
and basically found out we have to reduce some
payloads in the inlet in order to have some
instability. And that is where we are, and we will
start to do our experiment very soon.
CHAIRMAN WALLIS: Thank you. Stole your
mic? I guess you'll have to speed up again.
MR. BAJOREK: Okay. The way I think I'm
going to do that is by not spending a whole lot time
on the rod bundle heat transfer project.
Joe talked this morning about the type of
data that we hope to get out of the facility. Just by
way of background, the facility itself is full height,
very well instrumented, essentially a 7 x 7 bundle
with the corner rods knocked out for a total of 45
rods.
The rods are protypical, not only in
length but also in diameter.
An interesting feature about the rods.
I'm sorry.
CHAIRMAN WALLIS: Well, I guess we've seen
a lot about this before. The question's always been
when we going to get some results?
MR. BAJOREK: The results we hope are
going to be coming in later in spring, early summer
this year.
As I mentioned, most of the work that had
been done at Penn State this year has basically been
in bundle construction, shaking down the facility,
putting in supply tanks, putting in the DC power
supplies. Their schedule right now is to begin
testing in, I believe, April of this year. They're
going to start with a battery of about 15 tests.
Those will be reflood tests and then continue further
on in the year looking at steam cooling tests and then
tests where there would be steam and droplets injected
at the bottom of the bundle.
CHAIRMAN WALLIS: My concern has been this
is an expensive long term program that someone's going
to cut the budget before it gives you any data at all.
MS. UHLE: We will be finishing our
reflood data, the first phase at any rate, by the end
of this calendar year.
MR. BAJOREK: It should be January.
MS. UHLE: Right.
CHAIRMAN WALLIS: It would be good,
though, to start showing some results as soon as you
can so that you can show that the program --
MR. BAJOREK: That's why I guess the
reflood tests are going to be up there first. Try to
get the most important information and then build
things later on.
MS. UHLE: We're also in the discussion
with Korea to extend the program with some of their
grid spacer designs in a collaborative effort to
extend the program to get even more data.
MEMBER SCHROCK: So does this imply that
pressurization problems will have been resolved?
MS. UHLE: Yes.
MEMBER SCHROCK: Yes.
MR. BAJOREK: Okay. The next facility
that I want to talk about is the ATLATS for the phase
separation. But kind of as a lead-in to that and the
problems that we're observing in the ATLATS, I want to
put that in light with what we're seeing from the
AP1000.
As I mentioned earlier, one of the big
differences in the AP1000 compared to the AP600 is the
very large increase in the total core power. We're
going to see much large superficial velocities at
anytime during the transients that we observe in the
AP1000.
They've changed the resistance of the ADS-
4 line, greatly reducing the resistance. They've also
reduced the resistance of the CMT. They made some
other changes to the PRHR. But primarily with respect
to entrainment processes, it's been the increase in
core power that's really going to drive things.
As I mentioned, we have been doing some
scaling evaluation from a top-down scaling
perspective. Westinghouse doesn't have too bad of a
story. Actually what they've done in the AP1000 by
increasing the ADS-4 valve, they've made it look
actually a little bit more like the SPES facility. So
when you look at the scaling parameters early on, it's
even better agreement with SPES, which they have used
for code validation, than the AP600.
Later on it still looks very much like
APEX in the OSU tests, not too far off. Critical
period where we're having some heartburn showing from
a top-down scaling perspective whether OSU is okay is
during this ADS-4 period. Part of the question comes
into what is the critical flow, as the pressure
decreases, what's the quality that leaves the ADS-4
line?
One of the issues that is definitely going
to be -- have to be taken up in the phase 3, however,
are items that come from a bottom-up scaling. Well a
bottom-up scaling looking at more localized processes
in the core, steam generator, where else in the
facility.
Where we see problems right now from the
scaling leads us again to look at phase separation at
the hot leg leading to the branch line.
These figures don't show up very well, but
the situation we feel that is going to become even
more important in the AP1000 than it was in the AP600
is this condition where we have a froth going up into
the hot leg and we're entraining some fraction of that
into the branch line and out. Now, this factors back
to the safety of the system and in the analysis,
because if we start to entrain large amounts of fluid,
you get a larger two phase pressure drop and that
delays the time at which you transition over -- excuse
me, drop to a low enough pressure that your IRWST can
begin to inject into the system. The question mark if
that period probe is too prolonged and you have too
high an entrainment, too high a boil off, you lead to
some part of core uncover.
The second question that is arising from
our bottom-up scaling is what goes on in this type of
a scenario where it might be a DVI line break? You
don't necessarily have a level pushed up into the hot
leg, but now we have a high-quality froth above the
core. The question is how much of that becomes
entrained in the gas flow, eventually up and out the
branch line.
Now, I think everybody remembers seeing
some of the test results in the ATLATS facility back
in July. And by way of background, the basic reason
for having the ATLATS arose from some of the AP600
beyond design basis tests that had been run which
showed that there was some core dryout when they
started with a lower inventory in the vessel. The
RELAP couldn't predict that.
We have a situation where we're showing
hints of core uncover, the codes aren't predicting it
and we know it was due, primarily, to not being able
to predict entrainment in RELAP.
Now, we saw the facility, saw some of the
some film clips and also saw how the facility behaved
in July. And the meeting, unfortunately or
fortunately, noted that, hey, there's some significant
problems with the facility.
First and foremost, they're system-
dependent oscillations. It seemed to make a very
large difference depending on whether they have
blocked off at the steam generator, have a line open
from the steam generator back to the top of the upper
plenum. Other comments that we have received at the
meeting is that there was an inappropriate use of the
previous data and existing correlations in their most
recent model development.
The model that they were developing seemed
to assume some type of an annular ring around the
bottom of the branch line. This was the physical
picture that was being used to develop a newer model
for onset branch line quality.
And, Dr. Schrock, you have made a number
of comments on the references that they were
incorrect. Previous comments have not been
incorporated. And that their use of this person's
data and this person's correlation was at least very
confusing, misleading and probably wrong.
CHAIRMAN WALLIS: This just doesn't just
affect the entrainment, it affects behavior of ADS-4.
In ADS-4, the choke flow, you sometimes see steam,
sometimes you see it very wet or even a slug of liquid
coming along, that changes the flow rate out of ADS-4
to the entrainment, and you can set up conciliatory
behavior. You may need to be able to model
conciliatory behavior, not just some average. I'm not
sure they're doing that.
MR. BAJOREK: No. What I want to go over
are the types of things that we've started to do since
that meeting. I'm sorry.
MEMBER SCHROCK: No, finish what you were
saying. I'm ready to comment further on the summary
of what we learned from that meeting in July.
MR. BAJOREK: Okay.
MEMBER SCHROCK: Are you done?
MR. BAJOREK: Well, I was going to go into
the things that we're going to be doing with the
facility and the things that we're going to try to do
--
MEMBER SCHROCK: In terms of the answering
Dr. Wallis.
MR. BAJOREK: I'm sorry, I'm not sure what
you're asking right now.
MEMBER SCHROCK: Well, what I'm trying to
make a comment on is this summary of the things that
you found to be significant problems in the old
facility as a result of the meeting in July. There
are, in my mind, important aspects of that that are
not reflected in this statement. One of them is the
fact that the code attempts to solve the problem by
saying it knows what the flow regime is. When the
flow regime is satisfied there's a potential for
entrainment of liquid in the branch line. For that to
be a reasonable proposition, you have to see that in
the experiment, in fact, you get stratified flow. In
that experiment you did not get stratified flow. You
had a sloshing back and forth.
MR. BAJOREK: That's certainly correct.
MEMBER SCHROCK: No stratified flow
evident, and therefore that's the number one question
I think to be addressed, is your problem in running
RELAP, TRAC -- whatever code it may be that you're
trying to do the calculation with it --is the
difficulty that you have the flow regime wrong or is
the difficulty something about the model that you use
if the flow regime is right?
MS. UHLE: Can I answer it?
MEMBER SCHROCK: So that you haven't
addressed that issue, and I think that's step number
one in coming to grips with how you're going to get
something out of the OSU facility that will solve your
problem.
MS. UHLE: Can I answer at this point, at
least, do you mind?
MR. BAJOREK: Go ahead. Go ahead.
MS. UHLE: Okay. I think what was shown
to you at the OSU facility was the goals of the
facility was to look at essentially each flow regime,
first starting with the horizontally stratified and
going into the intermittent regimes as well. And what
the movie that was shown to you was looking more at
the intermittent. And we have data from --
MEMBER SCHROCK: Well, we weren't looking
at a movie, we were watching what was happening in the
facility.
MS. UHLE: Well, okay. I wasn't there. So
you saw not a movie, but the real facility.
That was for an intermittent regime. We
do run in horizontally stratified mode. And there was
date taken for the horizontally stratified. I don't
think that was communicated to you because it was
Research's goal to develop phase separation models
spanning all flow regime and horizonal pipes.
And the reason why we looked at -- or the
first attempt was to see if we could, regardless of
flow regime, come up with a correlation that just
looked at, say, average level and superficial
velocities was because of exactly what you're saying.
That the code, if you took this model and you applied
it across all different flow regimes, the answer you
would get would be dependent on what code you're using
and its prediction of flow regime. And so they tried
to come up with factors such as average level
superficial velocities regardless of flow regime to
come up with your entrainment rate. And that didn't
pan out.
So what will happen now is we will,
unfortunately, have to rely on the fact that you know
your flow regime and take the data and make sure that
we are consistently determining the flow regime for
the horizontally stratified case as well as the
intermittent, the wavy.
MEMBER SCHROCK: The flow regime that we
saw was -- the flow condition which is not described
by the flow regime maps.
MR. BAJOREK: I think your basic question
is okay, you have this condition in the facility. The
flow regime maps and the code right now, and for at
least the next several years, are static. They cannot
track waves or track the development and change of one
flow regime down a pipe. In the long run, we would
hope we would hope that we would get that type of
thing out of the Purdue or, actually more appropriate,
the interfacial area transport being done out of
University of Wisconsin.
For right now we're kind of stuck with
regimes we have.
MS. UHLE: And they are looking at the
fact that they were getting reflection from the steam
generator side and getting rid of that to run the
intermittent tests, you know, getting a flow regime
that is not --
CHAIRMAN WALLIS: It maybe required that
you ask them to develop a general correlation
entrainment out of the branch but using a geometry
that looks like AP1000. Because if you had a long
pipe instead of just try to find a flow in the other
branch, you might get something completely different
at what you see. Maybe you ought to be focusing more
on what actually happens in something which simulates
AP1000, therefore results might be at least useful for
analyzing AP1000. Don't claim this is some sort of
scientific study of a branch pipe under other
conditions.
MS. UHLE: But it's not just protypically
AP1000. I mean, in some ways PWR or hot leg
pressurizer. I mean, there are a few LODs down from
the --
CHAIRMAN WALLIS: How many LODs do you
actually get? Maybe you could analyze that, what will
actually really happen, get a good correlation in
terms of models of what really happened and not try to
mix in almost for something else, with 1000 --
MS. UHLE: But in a reactor, I mean you're
typically more fully developed. So the point of having
a really long, horizontally stratified regime, there's
no place in the reactor that you ever would be fully
developed.
CHAIRMAN WALLIS: Right.
MS. UHLE: Right. So we're trying to
identify the horizontally stratified in a sense that
as horizontally stratified as you can get in a
protypic reactor geometry. You know, that's the hard
part.
MEMBER SCHROCK: Well, what we saw in
Oregon was explained to us as the experiment that was
used to produce conclusions about the containment
problem and in fact the level shown as a level which
is determined by reducing the flow rate until it
ceased to have liquid entrainment.
MS. UHLE: And then they came back up in
the other direction and it mismatched.
MEMBER SCHROCK: I didn't see any coming
back up in the other direction. I asked about it and
there was no answer at that time. Maybe those came
after that question was asked; I'm not sure. But I
think -- what I'm reading here doesn't convince me
that you have a clear picture yet of sorting out what
you're going to have to do to get useful information
out of the OSU experiment.
CHAIRMAN WALLIS: I think that will
probably be in the conclusion. Sometime we're going
to summarize.
MR. BAJOREK: Well, let me go through and
summarize the actions as we see them right now.
CHAIRMAN WALLIS: The pictures you're
showing here of these double bumps -- that's not what
happened?
MR. BAJOREK: No. No. No, that's --
CHAIRMAN WALLIS: That's a fantasy?
MR. BAJOREK: No. And I mean it certainly
wasn't what we saw in the facility. But we think that
the goal one on this --
CHAIRMAN WALLIS: It's an analyzable
situation, but not relevant.
MR. BAJOREK: -- We think the first thing
is the basic, we're going to try to better understand
the system oscillations. The question is these
oscillations as we see in ATLATS facility, do they
also occur in the APEX facility, and how transferrable
is the information that we're getting from ATLATS to
the full-scale AP1000? The scaling that was done for
ATLATS, as well as for the OSU hot leg and branch
line, were based upon having the right void fraction
in the upper plenum, the right L over D between the
upper plenum and the branch line, and the correct
capital D to small d ratio between the branch line.
It really did not look at anything on the length
between the branch line and the steam generator, size
and heights of the waves that might form in a pipe of
diameter D.
What we would like to do is to try to
understand that better to realize whether the waves as
we see them at ATLATS are also going to occur in the
larger scale facility. That would be a review of the
scaling criteria.
Now, we did take Dr. Wallis' suggestion
and asked them to run a series of tests in which they
injected into the top and this figure shows what had
been intended to be porous injectors to go into the
core -- it doesn't show up very well at all. But it
does have an auxiliary air port by which we an do
injection into the top of the facility.
They ran those tests, they sent those to
us earlier in the month. We haven't had a chance to
go through those in great amount of detail, however my
observation in taking a look and plotting the liquid
levels, there still are a very high amount of
oscillations. It does not appear as to whether that
smoothed things out.
We've since gone back to them and asked us
based on those oscillations what are some of your
averaging procedures? Because we see a lumping of
much of this data, some of which Jennifer noted was
horizontal stratified. If we had those movies, I
think we'd be able to see that. A lot of it was
intermittent.
We think at this point we need to start
segregating that data into information which was truly
horizontal stratified and something that it is
intermittent, wavy, what other type of flow regime
that was apparent.
In future work we're going to request that
they supply a CCD or some other recording of what that
flow pattern was. Our expectation and understanding
going into the meeting is that we were going to see a
lot of horizonal stratified flow. We'd like to try to
get that recorded in addition to the comments that
they do have in the test reports. You have to dig to
find them, but there is a visual observation on what
that is. We'd like to start with that and segregate
out the points and get them into their appropriate
regimes right now.
CHAIRMAN WALLIS: Just want to make a
point that joining of the hot leg and the vessel is
accomplished by intersecting two cylinders leaving you
with a sort of strange, sharp edge around that
opening. That doesn't exist in the plant. It could
be of some significance in the hydrodynamics that
you're looking at here.
MR. BAJOREK: Okay. With respect to the
references and their use of the data, we've also asked
them to supply all copies of the references that
they've been using in assimilating their report and
plan to ask them to rewrite that section where they
talked about their literature search.
Based on the information we got, we agree
with you that it's confusing and misleading the way
things have been lumped together. Some of the reports
are difficult to get. We've asked that they supply
them to us. We're going to do our own review.
We feel that the model development needs
to be revised to be more regime-dependent. If we can
some day lump everything together, that would be the
simplest thing for the code application. But based on
what we've seen, we should use the horizontal
stratified data, keep that with models which are
appropriate for horizontal stratified flow regimes.
This in your ring picture that seems to
have been used in their development certainly didn't
show up in the experiments. And what we would prefer
is, rather than this figure over on the right which
they assumed, go to a picture which a similar model
had been devised by, I think it was Yanamoto, where
the picture, physical picture of the fluid beneath the
branch line is something that is forming more a
conical or a pyramid shape. At least that physical
picture looks -- corresponds much closer to what we --
CHAIRMAN WALLIS: But if you have a photo
that looks like a sketch?
MR. BAJOREK: There is one in the report.
CHAIRMAN WALLIS: We found a situation
where at least at the moment it looked like that
picture.
MR. BAJOREK: In fact, it was interesting
how they did it. They must have had a boroscope
inside the pipe looking axially. And you can see
almost a formation of a water spout.
CHAIRMAN WALLIS: Well, maybe it does for
some regimes, but what we remember very much was
large-scale oscillation of the whole pipe.
MR. BAJOREK: Yes. Yes. In fact, the
oscillation was between the branch line and the steam
generator.
CHAIRMAN WALLIS: Okay. We have to move
on to the next one.
MR. BAJOREK: Okay.
CHAIRMAN WALLIS: And then you can 15
minutes of summing up or so.
MR. BAJOREK: Okay. Subcooled boiling is
work that is going on at UCLA. We had Professor Dhir
come and present his results to us about 3 or 4 weeks
ago.
Now, the work that's being done at UCLA is
also in response to AP600 and AP1000, where there's a
realization that most of the decay heat removal is now
going to be done at lower pressure. We feel that the
models for subcooled boiling would not be as good at
the lower pressures as they were at higher pressures,
typically where you would need them for small break
LOCA.
The other question that's going to be
answered by the UCLA work is the idea of heat flux
splitting. How much of the energy and subcooled
boiling goes into void generation versus sensible
heating of the liquid? Right now whether it be in
TRAC or RELAP, the models are largely ad hoc. Based
on some limited test data to come up with the models,
but nothing very mechanistic in the way its treatment
of this heat flux split. So the objective of the UCLA
work, very much like the Penn State for dispersal of
film boiling, used advanced instrumentation and
detailed facilities to get high-quality information by
which we can develop these mechanistic models.
MEMBER SCHROCK: I don't know if you read
the comments that I made in a recent report, but my
recollection was in reviewing the -- TRAC's
documentation back in 1987 was that they had heat
transfer directly from the wall to the vapor, and
cases where the wall was -- the flow regime map let
it. So the transfer is nonphysical. It has to be to
the liquid and then to the vapor.
So I think it would be useful if this
could get sorted out in the way this was being
described and state more clearly what this heat flux
splitting is all about, and the way it's used in the
code. It originated the subcooled boiling models and
that got massaged and massaged and massaged and came
out as a GE thing under Leahy's name. But I think
that a lot of people have had confusion about what
this heat flux splitting means. And I think most of
those people have been people who work with codes, not
the people with experience. It needs to be dealt
with.
MR. BAJOREK: Well, I think there's the
physical models, and I think what you're pointing to
there's -- a lot of times the code -- you look at the
hydrodynamics and you pick your flow regime on one set
of conditions and then you take the wall temperatures
and maybe some gross estimate of a void and say this
is what's going on near the wall. But those physical
pictures may not necessary correspond. You may
predict a bubbly flow, who knows whether the bubbles
are concentrated out in the fluid or close to the
wall. Those selections have to be consistent --
MEMBER SCHROCK: For heat to be
transferred from the wall to the vapor, you have to
have a dry wall.
MS. UHLE: Right. But see, we have this
problem --
MEMBER SCHROCK: In the physical work.
MS. UHLE: But in the problem of numerics
if we're taking one second time steps, you can't do
that because you would get way too much super heating
of the liquid. And you can't get around that. So
that's where some measure of realizing that you're in
a numerical system in some way differs from reality,
and that's why you have to rely on assessment.
MR. KELLY: May I make interjection? This
is Joe Kelly from Research.
What Professor Schrock is alluding to goes
way back, in forced convection flow they did things
like void fraction volume --
MS. UHLE: Not necessarily. That's what
I'm trying to say is the fact that if you're taking
over a period of a time step of a second, you can't
put all of the liquid, all of the heat flux into the
liquid. And you know currently in our numeric systems
or numeric schemes and we would have too much super
heat the liquid and then the next time step you would
get the interfacial heat transfer to the vapor. I
mean, that's all I'm trying to say is that whether--
how we solve this problem we can talk about later, but
why it was done in the past, it may sound weird to you
but a lot of it is simply because we had to work in
the numerics of the time. Now when we make the code
more implicit, then we can get rid of those things.
But the computer limitations in the past prevented
that because we just didn't have enough memory or
speed of the computers were too slow.
CHAIRMAN WALLIS: I think that again we're
getting into too much detail. If we're going to
review Professor Dhir's work, we're going to have to
spend a whole afternoon.
MR. BAJOREK: Well, it would take quite a
bit of time.
CHAIRMAN WALLIS: There's no way that we
can get an overview of these programs beginning at
that level. What I get from this is that there is a
problem with predicting the amount of voids you get
and the heat flux in subcooled boiling --
MR. BAJOREK: Yes.
CHAIRMAN WALLIS: And that probably
sometime during the year we may need to look at this
in more detail.
MR. BAJOREK: Yes. I think sometime in
probably the spring would be the right time.
CHAIRMAN WALLIS: I don't recall actually
having a presentation.
MS. UHLE: It came to the staff.
CHAIRMAN WALLIS: Well, maybe that's where
we could contribute.
MR. BAJOREK: To summarize what he is
working on. Breaking up the wall into several
components using high speed visualizations. Two
different test sections, one a rod bundle another a
flat plate test section. Flat plate in order to give
him things like nucleation site densities, motion of
the bubbles, collapse rate of the bubbles as they
leave the wall and then getting additional information
from the rod bundle to augment that.
I've left in the package the types of
measurements that are being obtained in the facility.
I'm going to jump to more conclusion,
closer to the conclusions.
He's been successful at developing a model
to predict the delta T at the onset of nucleate
boiling in a subcooled flow, that seems to do a pretty
decent job at predicting not only his own data, but a
fairly substantial set of data that he also obtained
in a literature search.
CHAIRMAN WALLIS: Your measure of success,
this sort of a picture?
MR. BAJOREK: Well, this would be one of
them. I mean, because he's trying to get the onset
correct. He's also trying to get the right heat flux
at the onset correct, simultaneously. Get the single
phase heat transfer coefficients and also be able to
get models for the bubble size and the rate of
collapse of those bubbles. It basically gives you the
condensation component of that split.
This shows the heat flux based on his
model to try to predict the heat flux, which is a
contribution of a partial subset of those terms, and
by in large it seems to do a successful job not only
of his data but also on other sets of data.
CHAIRMAN WALLIS: Could you give us time
to see what the state of the art was before he came
along, and if he drew a picture like this based on
whatever you were using before, is this an
improvement?
MR. BAJOREK: Yes, it would. I mean, he's
also done the comparisons to some previous models and
you can see where the scatter is significantly larger.
MS. UHLE: Joe Kelly has a good paper on
the subcooled boiling model if you'd like to see the
current state of the art.
MR. BAJOREK: Okay. Now where he's going
with this work now, he's gotten enough data on the
flat plate test section. Most of the work that's
going to be done in 2002 is to try to come up with a
better term for this flux split, to get the additional
terms in this heat flux contribution to the total heat
flux during subcooled boiling, expand the data base,
getting additional information for the rod bundle,
increase the range of subcooling and look at some
higher pressures. Right now everything is fairly
close to atmospheric.
I think he can take the facility up to
close to 3 or 5 bars. And he thinks that most of that
work can be completed in 2002, which is why in the
overall schedule we're looking at trying to implement
those models later in 2002, but probably not in time
to get into the Rev 0.1 release.
Okay. The last topic that we're going to
have is looking at the interfacial transport, which
has been done primarily at Purdue and University of
Wisconsin. Jennifer's going to talk about that.
MR. BOEHNERT: Do we have these slides,
Jennifer?
MS. UHLE: No, I'll get them. Because I
thought that Steve had them, he thought I had them.
I've talked about this before at the ACRS
meeting, so the objective of the interfacial area work
is to get away from the static flow regime use in the
code, the reason being is that for one thing, we need
to use interfacial area in the code. It's a value that
we have to have a closure relation for. It determines
the interfacial heat transfer as well as the
interfacial drag.
We currently model it using static flow
regimes. I think everyone's aware of the fact that the
flow regimes were developed in steady state
situations, lots of air/water. At any rate, the
transition criteria and the use of the static flow
regime, it doesn't represent the actual physical
processes of creation and the destruction of the
interfacial area, so there's no time and length scale.
And so if you change your flow rate; in some sense if
you have an oscillation you instantaneously change
your flow regime. That doesn't sound that bad, but if
you consider the situation of annular flow and you
increase your vapor flow rate so that it's beyond the
point where you're entraining liquid drops, you can
increase by several orders of magnitude the drag in
the interfacial area of interfacial heat transfer so
that it causes this oscillation and it can also cause
some inaccurate answers.
So we're trying to develop a first order
equation for the transport of the interfacial area;
that is the objective. We realize --
MEMBER KRESS: If you have that, you no
longer need flow regimes at all?
MS. UHLE: That is the goal, but before we
can take out the flow regimes we have to have the data
in the model covering all flow regimes in geometries
prototypical of nuclear power plants. And that is the
big effort. That's why we're trying to collaborate
with France and Japan, and open this up for
international collaboration, so it's a big scope. We
are making progress ourselves but we realize that we
had originally thought that Japan and France were
going to provide the steam water. We're still working
on that.
MEMBER KRESS: What is the position now?
MS. UHLE: I can go through where we currently are.
I just want to point out, though, if we don't to the
point where we actually do use this interfacial area
transport equation for the flow regimes, this project
is not useless by any stretch of the imagination
because of the fact that we do use values for
interfacial area.
We will be able to take these measured
quantities of interfacial area and then compare them
to the correlations that we use in the code currently
to make sure that we are at least getting a prototypic
value of interfacial area for the flow regime of
interest.
So, again, if the modeling doesn't work
out in the long run, the data is still useful.
CHAIRMAN WALLIS: This principle has a
separate conservation equation --
MS. UHLE: Yes.
CHAIRMAN WALLIS: It's something you can
could put into your TRAC now --
MS. UHLE: Yes, I did that. Yes, I did
that. If you remember a couple of years ago where I
put in the first or I put in the -- one group
interfacial area equation. In other words, it was for
bubbly flow. So by one group I mean that the vapor
phase was all spherical, and therefore the drag
coefficients were, again, first spherical
configuration. And in --
CHAIRMAN WALLIS: So how long did it take
you to do -- a long time?
MS. UHLE: Yes, it took me a week. It
took me a week, and that includes modeling and
comparing to the data, although I did call our
numerics guru for a few challenges along the way.
CHAIRMAN WALLIS: You called this TRAC-M
development.
MS. UHLE: Yes.
CHAIRMAN WALLIS: And the thing is if
these guys are successful --
MS. UHLE: It'll go in easily.
CHAIRMAN WALLIS: -- in 2004 or something,
put into TRAC as an implement?
MS. UHLE: Yes.
CHAIRMAN WALLIS: With an option or
something.
MS. UHLE: Right. Now with the two group,
it's going to take me more than a week. I was going
to do it this year, but then I said I was demoted to
assistant branch chief and they don't let me touch the
code anymore. But I was planning on doing that to put
in the two group equation. And there's a little bit
more complexity with the two group equation, because
you do have to solve a matrix.
CHAIRMAN WALLIS: All you have to do is
delegate somebody younger and quicker.
MS. UHLE: I thought it was older and
wiser.
PARTICIPANT: That's his answer why we
hire lower grades --
CHAIRMAN WALLIS: I'm not sure we need to
spend on this. It's going on it's processing --
MS. UHLE: It's going on. We're covering
flow regimes. With respect to Professor Schrock's
questions, we've covered bubbly flow -- sorry. For
the co-current upflow we've covered all flow regimes
up to annular. We're doing counter current flow.
We've completed bubbly flow. Started to do co-current
down flow for bubbly flow. Horizonal, we've covered
all co-current regimes and we're starting to extend to
other geometries, so we do have this database and
comparing.
There are two group models that they've
come up with, although it's not put in the code, they
do compare to the data as they develop it.
In the future, we need to go to steam
water for the source and sink term of the phase
change. And, again, extending to just other flow
regimes and geometries.
We're hoping to have the final model, you
know, our ideal would be to have it in 2005 in the
code and replace the static flow regime. It depends
a lot on --
CHAIRMAN WALLIS: I hope that they are
publishing results --
MS. UHLE: Yes. Yes. You haven't been
reading International Journal of Heat and Loss
Transfer then because, yes, we just published
something.
CHAIRMAN WALLIS: You're jumping ahead
with that accusation. You don't know what I've been
reading.
MS. UHLE: Yes, we've been publishing.
Are we done except for the summary?
CHAIRMAN WALLIS: Well, I was hoping that
we could talk about the papers
MS. UHLE: I can give you the papers, if
you'd like.
MR. BOEHNERT: Yes, we'd like the papers.
CHAIRMAN WALLIS: I was trying to jot down
as where we could interact with you in the future that
would be profitable. My colleagues should come in on
this. But I feel with the development of this
consolidated code that what we should do is encourage
you to keep up your enthusiasm for the activity but I
think where we might contribute is in the
documentation. Do you have draft documentation that
we can look at and give you some input and avoid
giving you surprises when we see it later on? Maybe
we'll make the documentation better? I think in the
other areas of Joe Kelly and company, doing work with
their former knowledge of what they're doing than we
are, I think that they go for it and we'd like to see
the result. But then I think we should discuss what
we need to do about each of the review of some of
these other programs, USU, Penn State. But can we
first look at other comments on the TRAC
MEMBER LEITCH: My question, I guess, or
comment is that I'm a little confused about what
release means. Does that mean that it can only be
used in certain circumstances? And if so what
purposes? In other words, are there other grounds or
to what extent -- seeing that we've talked about
perhaps 8 or 10 applications and I guess what I think
should be the outcome of the status of this by the end
of next year. It would be useful to address these
other applications. What will the status be? In
other words, I guess we've seen a program that ends
sometime at least in the next 13 months, but obviously
the research effort is geared towards the targeted
applications. But I'm just a little confused as to
what will be value of using the TRAC code and the
RELAP5 in these applications. I guess that's a
reasonable question.
MS. UHLE: No, no. That's a very
reasonable question, and it's a quick answer here. Is
that by the end of 2002 we will be as good as the old
codes for the targeted applications, so we can from
then on rely on the TRAC code.
Now, the fact that RELAP5 is now the
workhorse code mostly for the international community
as well as for NRR, we foresee bringing that in-house
and maintaining and it using it as a benchmarking tool
as we continue the effort. So it's not like we'll be
dumping RELAP5. But at that point we'll be able to
use either code for, say, the PWR applications. We
will, of course, think that TRAC will be better for
the large break LOCA for the PWR.
We will be as good as TRAC-B used to be
for the BWR applications. And we can do stability and
3-D kinetics for the BWR to replace RAMONA.
So, again, we'll be starting to focus and
start this transition into relying on TRAC-M.
MEMBER LEITCH: The synergistic effects?
MS. UHLE: That will be done with TRAC-M.
MEMBER LEITCH: And the PBMR?
MS. UHLE: TRAC-M. Right. But we will by
the next time we -- we say the next fall meeting, we
will have the physical models in to do the PBMR.
Hopefully, have identified data sources and at least
have a few plots to show with respect to system
behavior.
MEMBER LEITCH: I guess just to
paraphrase, I think what I heard you saying is by the
end of next year this is when it comes out, TRAC-M
will be equal to or better than?
MS. UHLE: Yes.
MEMBER LEITCH: So you will make another
presentation to the committee?
MS. UHLE: Yes. That is the goal, yes.
CHAIRMAN WALLIS: So you will come to us
towards the end of next year with a consolidation of
the codes?
MS. UHLE: Yes.
CHAIRMAN WALLIS: You will then have
consolidated the codes?
MS. UHLE: Yes.
CHAIRMAN WALLIS: Right so you won't have
improved them much.
MS. UHLE: In some cases, for example, the
level tracking we've improved. The large break
calculations with Joe Kelly and Weidong Wang's reflood
work, we would have improved. Hopefully we will have
the phase separation stratified flow model in for use
with the AP1000, we would have improved that.
The other improvements have been more user
convenience , speed, robustness rather than physical
models. And then at that point in time as we then go
into more of a PIRT base developmental assessment
effort and continue working more closely with the test
programs, we would then focus on improving the
physics. But the original charter of the
consolidation and what the Commission had signed off
on was to recover capabilities by the end of this
period, and we feel will achieve that.
MEMBER SCHROCK: Do you have any
indication that industry wants to start using it?
MS. UHLE: Not so much industry.
Industry's interested in the graphical user interface
because it works with RELAP5. And, of course, there's
the strong use of RELAP5 in industry.
Shanlai Lu on the staff has for NRR's use
has taken the TRAC-G and developed a pearl strip that
allows us to take a TRAC-G input deck and convert it
into what TRAC-M can run. So NRR would be using that
and is a comparison for the future application of the
TRAC-G submittal for the large break case.
The Naval Reactors is very interested in
using the consolidated code because Betest and Capital
are looking at, and in fact consolidating their
analytical work as well.
And then, of course, the international
user group is holding off, you know, waiting to see
how it works. Most people are interested in the
ability to recover RELAP5 functionality, and that is
what we need to -- we will proving this in a month or
so.
MEMBER FORD: I've got three comments.
Not being a fellow hydraulic person and not specific
to the physics.
First is, you weren't clear about the
qualification of the code especially when you're
qualifying it against scattered databases. Presumably
the code should be able to predict the uncertainty
that you have.
The second question, and actually more a
comment. The second question is what will the
hierarchy be for the various codes when this TRAC-M
code versus the licensee's code, what determines
whether one is better than the other? Really a
professional comment.
The third one is really also a comment. In
the beginning that mission statement said safety
margins and therefore presumably the next stage after
TRAC-M is to incorporate it into materials
degradation, and I'd be interested to hear about that.
Is that your ultimate goal?
MS. UHLE: I'm Sorry.
MEMBER FORD: Well, aging phenomena of the
materials.
MS. UHLE: I mean, our work with the
materials interaction really is coming from providing
thermal-hydraulic conditions to the division of
engineering, and as well as working with the PRA
branches where we provide, you know, based on the
material degradation at these new thermal-hydraulic
conditions, or maybe because of flow induced or flow
accelerate corrosion you're going to have a higher
failure -- or sorry. A higher break frequency, you
know, that would go into the PRA. You know, that's
more of our interaction.
MEMBER FORD: Well, there's a lot of
material degradation issues when you have a synergy.
Presumably that's all been passed along --
MS. UHLE: Typically the level of detail
you need to couple thermal-hydraulics to something
like flow accelerated corrosion is not going to come
out of a system code, because our nodes are like this
big. And you're looking at the boundary layer to look
at, you know, the physical processes going on to do
the flow accelerated corrosion. And more --
MEMBER FORD: I see.
MS. UHLE: That would be more a
computational fluid dynamics linkage.
MEMBER FORD: Perhaps this phenomena is
related to the core shroud.
MS. UHLE: Again, the idea of the thermal
fatigue cycling, that again is looking more at large
eddy simulation to get the frequency of the water
coming up at a different temperature, going back down.
That is not something a system code is ideally suited
for. That would be more of a computation fluid
dynamics application.
And we do have CFD technology in-house,
and we are, again, hiring to increase that and that is
something that we can think of as far as interacting
with the division of engineering as these applications
come up.
CHAIRMAN WALLIS: What I had in mind in
this summary was we give you some input, and we speak
again about more activities, say, in six months, and
how we can interact in the next six months.
Tom, did you have -- I think we're talking
about the TRAC-M --
MEMBER KRESS: Yes. TRAC-M. I did have
a couple comments, but I'm not sure my comments are on
how best to interact. My comments are more with
respect to what Ms. Uhle was saying. I think you
ought to view integral experiments as rough. I don't
think you're going to predict experimental error.
We're talking about two different things. Go look and
see if your predictions fall within the boundaries of
the experimental error. So my comment there is use
separate effects testing to determine the
uncertainties in your specific model. Think very hard
about how to incorporate them in the code in a way
that you get an uncertainty distribution in your final
product. When you get to that point, you really have
a code that is very useful.
My other comment I had is that I certainly
like what I see and I encourage you to continue with
this. If you are very successful it would solve a
whole lot of these problems with flow regimes, how you
transition from one to the other and how you deal with
them on the code. I'd certainly like to hear more
about that later.
CHAIRMAN WALLIS: Can we move on then to
the separate programs?
The OSU program, I'm not sure -- any hope
of bringing them around to our viewpoints? Why don't
we try to figure out if there is some way in which we
can interact. I don't want to be with you or them at
the end of the program and have exactly the same
comments we had when we visited.
MS. UHLE: As part of my action items that
I have written down, it is to schedule some sort of
test program review if I can interact with Paul to do
that. Because, obviously, with your expertise it's of
good value to us to learn.
CHAIRMAN WALLIS: Right.
MS. UHLE: And I mean I think our goals
our consistent really, although sometimes it can be a
combative interaction. You know, we want an accurate
code, we want to be able to extend the code to other
applications easily, we want to be able to understand
uncertainty and calculate it so that this tool can be
of use. You're looking at a whole lot of people that
have put a lot of time in this program, and the idea
of it not being useful, you know, we wouldn't get out
of bed in the morning.
So I do think that our goals are
consistent, and so I think further interaction with
you on a more frequent basis can only benefit us.
CHAIRMAN WALLIS: Maybe there is a way in
which OSU can come before this Committee before the
report to the full assembly. There's no way.
MS. UHLE: That's --
MR. BAJOREK: No, that was for the ETS.
CHAIRMAN WALLIS: ETS. The other work is
still going on?
MR. BAJOREK: The work is --
MS. UHLE: Oh, yes, we're getting a
preliminary model for the --
CHAIRMAN WALLIS: Work on a useful
interaction with OSU that we could have.
MS. UHLE: Yes. Okay.
CHAIRMAN WALLIS: Penn State, it seems
they're still building the apparatus, they haven't
gotten their results. I'm not sure we have anything we
can --
MS. UHLE: I mean, they're doing shakedown
testing now and characterizing like volumes and lost
coefficients and things of that nature.
CHAIRMAN WALLIS: And PUMA oscillations,
I don't think we have anything to get until they start
doing something? We might, I think, contribute to the
critical flow models.
MS. UHLE: Right. I have down to give you
that critical flow report for Professor Schrock.
CHAIRMAN WALLIS: And maybe you can evolve
at some time an actual presentation by them?
MS. UHLE: Yes.
CHAIRMAN WALLIS: And then on DEER, I
think we really are due a presentation. It's been
going on for some time, we have not had the detailed
interaction, the kind of questions we'd love to ask
and don't have time for today, so maybe we should
schedule something for later. After the start of the
year maybe.
MR. BOEHNERT: That'll be fine.
CHAIRMAN WALLIS: All right. Anything
else on -- I didn't have anything immediate
MR. ROSENTHAL: You expressed an interest
in some of the MOX work.
MS. UHLE: MOX work.
CHAIRMAN WALLIS: There is a fuel
subcommittee of the ACRS.
MR. ROSENTHAL: Yes, I think that will be
a better place for that and we could advise it --
CHAIRMAN WALLIS: It's all from a drop,
it's neutronics.
MR. ROSENTHAL: We would do the neutronics
and other MOX related issues about how you load the
power and then the source.
CHAIRMAN WALLIS: But we'll do that with
kind of a separate subcommittee on MOX. Maybe that's
where it actually --
MS. UHLE: Okay. Yes, I have that down.
Although we will give you some written information to
respond to Professor Schrock's questions. Although it
may not be answering all the questions that you've
asked, that can be at a future time.
CHAIRMAN WALLIS: I think that the purpose
of this meeting is for us to give some input for the
main committee and the writing of the research --
MR. BOEHNERT: That's correct.
CHAIRMAN WALLIS: This does not require
some letter or anything?
MR. BOEHNERT: No, it does not require a
letter. It's fed into the work on the research --
MS. UHLE: I mean, one thing I do want to
point out, because based on the feedback you give us
annually is that we understand we need to tie in our
test programs closer. It's not news to us to hear
that.
In the past it's simply been how much time
we have and the staff we had available. Now that
we've been in this hiring mode and we've been bringing
more expertise in-house, we're going to try to start
to do that. It's a big focus for us.
CHAIRMAN WALLIS: I just want to be
encouraging.
MS. UHLE: We realize that, it's not -- in
fact, you know, Steve being the senior level scientist
here is the perfect person to really lead that
initiative, and he's been doing a great job in trying
to tie in the model development work, the test
programs more closer to the code development work that
Joe Kelly will be overseeing.
CHAIRMAN WALLIS: I hate to go back to the
TRAC thing, but you know you have done a good job of
consolidating these codes and at the end of this next
year you're going to show that they'll at least do all
the things the previous codes did, which is a bit like
saying Amtraks going to run at least as fast as the
steam trains used to run in the '30s. And what we're
really looking forward to is that there's a high speed
train or something that's really that much better.
That's what we'd love to see.
MS. UHLE: Right.
CHAIRMAN WALLIS: The sooner that can get
on the track, the better.
MS. UHLE: Yes, right. Again, we agree
with that. Now, we've been doing what management has
assigned us to do as far as the Commission policy
being to do the consolidation first before we start
the improvements. We also want to make improvements,
probably faster than you do, because we're the ones
doing the work and it frustrates us more than it
probably frustrates you. So that is going to be our
focus.
But we were tasked with this consolidation
effort first, and that was the high priority.
CHAIRMAN WALLIS: Even if you have to
smuggle the improvements along.
MS. UHLE: We've done that.
MR. ROSENTHAL: No, no. But we work in
accordance with the operating plan, of course. But,
no, in conjunction with the synergy work we have
planned ATWS calculation and we can now do couples,
3-D, space time kinetics, really better ATWS
calculations than we were able to do. And that'll be
a shorter term product. But we do some benchmark.
So we're going to start seeing the
benefits now.
MS. UHLE: Ready for your next victim? I
think they're behind you.
CHAIRMAN WALLIS: I'm very glad that you
have all these people now to work on these problems.
It's good to see Joe Kelly back here. Go away with a
good feeling.
MS. UHLE: I also went away with the
concept that any PIRT that you are involved in we're
supposed to ignore, because you are unduly prejudice,
that's the number one lesson we learned today.
CHAIRMAN WALLIS: I think I'm utterly
clean. I don't think I've ever been involved in a
PIRT. Now nor have I ever been.
Do you have a few final remarks? Then we
will close this part of the meeting.
I would like to take a break. I notice
there are all the people waiting. We have caught up
some time, so we'll try to keep on time, at least get
out of here before 6:00.
We'll take a break.
(Whereupon, at 2:43 off the record until
2:02 p.m.)
CHAIRMAN WALLIS: No introduction, Mr.
Henry, please begin.
MR. HENRY: Thank you, Mr. Chairman.
We're happy to be here today just to have
the opportunity to present to you this new activity of
using the MAAP5 containment code to replace the models
for containment integrity at both Beaver Valley and
Point Beach.
Before I get into talking about it, I
thought perhaps you would like to hear from the two
different sites of their motivation for going to a
different code for containment integrity, that it has
some differences which are site specific. So, maybe
just a couple of minutes with each site.
I'd like to introduce Mike Testa from the
Beaver Valley site and then he'll be followed by Harv
Hanneman from Point Beach.
MR. TESTA: My name is Mike Testa. I'm
with Fist Energy, and we operate the Beaver Valley
Power Plants, that's Beaver Valley 1 and 2. And I'm
Project Manager for the Power Uprate that's being
undertaken there. The power uprate that we're looking
at for the Beaver Valley plants to increase the power
in total by about 9.4 percent.
The MAAP code and the use of the MAAP code
is an integral part of that, and I just want to give
you, as Bob said, a minute or two perspective on the
use of MAAP at Beaver Valley.
The Beaver Valley plants are three loop
Westinghouse PWRs. The architect engineer was Stone
and Webster. The containments were designed
subatmospheric, that's the way they're currently
operated. And we want to use the MAAP5 computer code,
basically, to replace the existing design basis
computer code LOCPIC. And using the MAAP5 code we
want to, again, reanalyze the containment and move to
an atmospheric containment.
Benefits for going to an atmospheric
containment are that right now for personal access to
the containment, it's in an oxygen deficient
environment and the people that access the containment
are required to wear supplemental breathing apparatus.
And this will eliminate the need for that. That goes
towards enhancing personnel safety on access to the
containment.
The other thing this does for us is that
with a move to atmospheric containment where we change
the initial condition for the containment operating
pressure, we're incorporating that into our best
estimate LOCA analysis. And this will allow us to gain
margin on peak clad temperature, so we'll be gaining
a benefit in that respect also.
And, as I mentioned, this supports our
power uprate initiative in that the power uprate is
going to be based on the containment analyses that's
done with MAAP5.
CHAIRMAN WALLIS: If you do not use MAAP5,
are you not able to get this 9.4 percent power uprate?
Is it critical?
MR. TESTA: Yes, it's critical in that,
yes, we've done some studies with the existing code
and with then we looked at MAAP5 and it affords us
additional benefit in that we can increase the initial
containment pressure and basically review or rerun the
design basis spectrum of accidents and stay within our
containment design pressure.
MR. BOEHNERT: You said you're going to a
best estimate LOCA code.
MR. TESTA: We're going to use that, yes.
Westinghouse best estimate LOCA.
MR. BOEHNERT: Westinghouse?
MR. TESTA: Yes.
MR. BOEHNERT: Have you checked with them?
MR. TESTA: Yes. Again, our plans for
MAAP5 is that we're going to utilize it for the
containment integrity evaluation. Again, we want to
replace the LOCPIC code. And in doing this we're going
to perform the analysis using MAAP5 code consistent
with the current design basis requirements, and that
we're going to analyze for LOCA, steam line break,
different spectrum of breaks and look at the
corresponding results, the response of the containment
given pressure temperature and so forth.
And the last thing is that, again, the
MAAP5 takes advantage of the latest experimental
information. And what we want to do with MAAP5 is
move or take the computer code in-house so that we can
put our engineers in a position to be able to utilize
the computer code and to make operating assessments.
We've been working up to this point with Dr. Henry to
develop the inputs and the parameter files, which is
a benefit to our developing our in-house expertise.
MEMBER KRESS: What was your last code
that you used before?
MR. TESTA: LOCPIC.
MEMBER KRESS: L-O-C?
MR. TESTA: Yes, L-0-C-P-I-C.
MR. BOEHNERT: When are you making these
submittals?
MR. TESTA: We talked about that yesterday
a little bit. There's going to be a topical submitted
in January time frame for the MAAP5 code and we're
looking at May for the Beaver Valley plant specific
submittal. And in there will be the MAAP5 code, the
analysis that was conducted, the results and also the
supporting information on allowing us to move to an
atmospheric containment.
MR. BOEHNERT: What about the LOCA code,
when are you going to make these submittals?
MR. TESTA: Well, the LOCA code, that will
follow. That will be later on in around September time
frame. And we're basically putting in the building
blocks for a power uprate submittal.
MEMBER KRESS: When you use MAAP5, and I
don't know if this is for you or somebody else, do you
use it differently? Do you use other sources also?
What do you use as the input?
MR. TESTA: The input of MAAP5 is going to
be the Westinghouse mass and energy input.
MEMBER KRESS: Okay. So you use the mass
and energy input?
MR. TESTA: Yes. Yes. Correct.
MEMBER KRESS: You'll only use the
containment part in MAAP5?
MR. TESTA: Right.
MEMBER KRESS: Then the NRC would use this
and they'd never have to check the containment?
MR. TESTA: Correct
MEMBER LEITCH: Does MAAP5 have the option
of one region, or five regions?
MR. TESTA: Do you mean as far as
analyzing or --
MEMBER LEITCH: Yes.
MR. TESTA: Yes. Right now the developed
model for Beaver Valley is 17 nodes for both Beaver
Valley 1 and 2. Basically the same model is broken
down into 17 nodes and review that for large breaks,
you know, which nodes or compartments they occur in
and then we're evaluating the corresponding response
within the given of the multi-node response capability
of the code.
CHAIRMAN WALLIS: So how many nodes are in
this MAAp5 code?
MR. TESTA: Seventeen.
CHAIRMAN WALLIS: How many in your present
code?
MR. TESTA: One.
CHAIRMAN WALLIS: One?
MR. TESTA: Yes.
MEMBER LEITCH: So to get the results
where the containment pressure is acceptable you not
only are changing the code but you're increasing the
number of regions analyzed.
MR. TESTA: Yes, that's part of what Dr.
Henry's discussion will be is on the benefits or the
need to incorporate multi-node model.
CHAIRMAN WALLIS: Thank you very much.
MR. TESTA: Thanks.
MR. HANNEMAN: Good afternoon. I'm Harv
Hanneman. I work for Nuclear Management Company and
the Power Uprate Project Manager for Point Beach
Nuclear Plant.
A little background, Point Beach is a two
unit site with two LOOP Westinghouse reactors, roughly
1500 megawatts thermal each. We have large dry
atmospheric containments for both units. And our
initial motivation for using the new MAAP5 methodology
is to support containment integrity analysis for
possible future power uprate of about 10 percent in
reactor power. And we're in the planning phases of
that project right now, and we saw the need to get
additional margin for our peak pressure and also
temperature in containment because of the 10 percent
higher reactor power.
However, other benefits that we expect to
achieve by the use of MAAP include the accommodating
a pre-accident containment pressure of 3 psig. So
that would be in our technical specifications in the
range of pressures that would be allowed in
containment initially.
Provides margin for some of the issues on
containment fan cooler service water boiling, which
came out of Generic Letter 96-06.
And also provides a plant specific main
steam line break containment analysis for Point Beach.
Currently our licensing basis is an
evaluation of a generic two LOOP Westinghouse analysis
for containment, so going to the uprate, we thought we
needed a plant specific analysis and we believe MAAP
will give us the margin that we need.
CHAIRMAN WALLIS: Does MAAP give you a
margin for this service water boiling issue? Does it
predict containment or something?
MR. HANNEMAN: We expect it to predict
slightly lower peak temperatures early in the
accident, and that's when boiling is an issue.
CHAIRMAN WALLIS: Coolant containment?
MR. HANNEMAN: Right. Right.
CHAIRMAN WALLIS: I understand now.
MR. HANNEMAN: So our application of MAAP
would be to use MAAP5 for the containment integrity
analysis for the plant. We would continue to use the
Westinghouse methodology for calculating the mass and
energy releases as an input for both LOCA and steam
line break accidents. We currently use the
Westinghouse COCO methodology for containment
integrity, and we would replace that with the MAAP5.
This would allow us to take advantage of
some of the latest experimental information that Bob
Henry will be discussing here in a few moments. And
it also provides us an opportunity to bring the
containment integrity analysis in-house so our own
engineering staff will be performing the plant
specific calculations; that'll give us greater
knowledge of that analysis in-house and also allow us
to perform more timely responses to any operational
emergent issues that come up with regard to
containment response.
MEMBER LEITCH: Harv, do you, like Beaver
Valley, also need to use MAAP5?
MR. HANNEMAN: We've done some initial
analysis using the COCO methodology for both LOCA and
steam line break. The LOCA peak pressure was slightly
under our containment design pressure of 60 pounds,
but the steam line break the pressure exceeded it at
the uprated condition. So, that's why we feel we need
this methodology to give us a little bit more margin.
MEMBER LEITCH: How many nodes are using?
MR. HANNEMAN: I'd have to defer to 10 --
9.
MEMBER LEITCH: Nine.
MR. HANNEMAN: Nine volumetric nodes. And
currently we have one also with the COCO.
MEMBER LEITCH: Just a quick aside to
Mike, the people at Point Beach are talking about an
initial pressure in pounds, do you have a similar
number for Beaver Valley?
MR. TESTA: Yes, and for the move to
atmospheric containment we're looking at developing an
operating band of 12 to 16 pounds. One atmosphere for
us is 14.3
MEMBER LEITCH: Thanks.
MR. HENRY: What I want to present for you
today is really a work in progress, and we
particularly wanted to get your feedback on the
approach. As you can see, there's a couple of sites,
that we really want to know how you feel about this
and what has to be done in the future. And has been
said by both of them, there will be a submittal to the
staff sometime planned early next year and it'll be
led by a submittal of the methodology itself for the
staff to begin to review.
But in addition to your feedback on the
methodology, as we go through this you'll see that
there's a lot of experiments here and the experiments
represent a level of understanding and the
capabilities of the calculational tool. If there's
some experiments that we haven't managed to cover here
that you think would be very helpful in understanding
the capabilities of the model, we also want to get
that particular feedback and get it early on so that
we can take advantage of the expertise on this
committee.
Obviously, I don't have to tell you. Feel
free to ask me any questions as we go through this.
But let me also say early on that there's obviously a
lot more material than we can cover in the time that's
allotted. I apologize for that. We'll go as slow as
you want to go, but we did want to bring to you the
fact that we've worked very hard at trying to make
sure that the model is comprehensive of these
experiments and in a very simple manner.
So, the things that I'd like to cover for
you today are the issues that are related to:
nodalization; representation of the atmospheric
motion, which is circulation within the atmosphere has
a major influence on the rate of energy transfer from
the containment atmosphere on a nodal basis to the
wall.
Let me also say just up front the
nodalization scheme and map is generalized. You can
have as many as you want to define. Right now the
code will allow you up into the range, of what, I
think 26 or so. But usually it's a very highly
compartmentalized containment that would need 26
nodes, but that's why there's different nodalization
schemes for the different plants. As an example, two
LOOP versus three LOOP gives us different
compartments. Different geometry gives represented
differently. But we'll talk about that as we go
through this.
CHAIRMAN WALLIS: Circulation is going on
within the nodes?
MR. HENRY: Within each individual node is
where it's evaluated, yes.
CHAIRMAN WALLIS: Now, I don't quite
understand that. So you have some sort of a model and
it interconnects the hose between the nodes, but it
also super imposes some kind of a circulation within
each node?
MR. HENRY: We will get to that. And the
place that's important, Graham, is that's what
dictates what the local boundary layer is and
therefore, the rate at which energy can be transferred
to the wall.
In addition to this, this blow down and
give you forced circulation flows, but then you also,
obviously, have to comprehend natural circulation
flows because there can be compartments that are
isolated or there's later in time when the flows die
down. Natural circulation dominates. That has to be
a key part of it.
Another very essential part, which is
nothing new to MAAP5, that's already in MAAP4, is the
ability to have countercurrent natural circulation if
you have heavy over light at an opening between the
two so that they can exchange mass and energy, and
that natural convection type phenomena.
Condensing heat transfer, of course, we
look at the condensing on cold heat sinks. We looked
very hard at the separate effects test, and that's
where our understanding comes from. And we try to
make that step to the containment analysis in a very
structured logical manner without any kind of games.
So our whole understanding comes from the separate
effects tests.
And then lastly, the influence of water
entrainment, and that's another place where the local
circulation velocity is important because we can have
water films on the walls, we could have water
accumulate on the floor. If you have velocity which
exceed the entrainment rate, then that material could
be picked up and put into the atmosphere.
So from our perspective, as has already
been discussed, we want to move to something from a
design basis approach to something which is more
realistic, and we hope to be very realistic of the
containment response.
And the issues that we see that are
involved in this are:
Certainly nodalization, because we want to
represent the containment geometry;
The need to represent the displacement of
noncondensible gases, and that's a major reason why
multi-node differs from single node because you can
displace air out of the region and, of course,
displace noncondensible gas means that for certain
conditions at short time frames the energy that's
transferred to the wall can be much greater. If you
have strictly a single node, then the partial pressure
of the air is always the same;
We need to represent the potential for
induced circulation, which means we solve the momentum
equation in the gaseous atmosphere as this blowdown
occurs;
And we want to represent the potential for
stratification, so we look at these nodalization
schemes. There's always a potential above the
operating deck of having more than one more node. So
if you have light gases, there is a potential that it
can accumulate in the top of the dome.
MEMBER KRESS: In read in the material we
received that there is no momentum equation in that.
Did I get that wrong?
MR. HENRY: I think so. But like I say,
there's a lot of momentum equations.
CHAIRMAN WALLIS: But they're not
transient momentum, they're in a pseudo-steady state.
MR. HENRY: Correct. Transient in the
sense that you give me the current conditions and
I'll--
CHAIRMAN WALLIS: It really should be
momentum DT in there.
MR. HENRY: Yes. Well, for a given cell
the momentum in a cell -- in a node changes given the
blowdown time. Circulation velocity is a function of
time. You can ask me the question when we get to it.
Maybe I'm misrepresenting or misconstruing what you're
saying.
CHAIRMAN WALLIS: Well, I think your
momentum equation does have a D by DT determinate. It
just balances.
MR. HENRY: Okay. Well, we'll get to it.
For the kind of nodalization schemes that we
recommend, certainly it's to move away from the single
node for reasons noted here, but you don't have to
have tremendous number of nodes. You just have to
represent the fact that the air can be moved to
different locations, that you can have stratification,
etcetera.
From our perspective, what it means to
have a realistic model, and I'm just going to discuss
both of these points together in time, to save us some
time.
We want to make sure that we represent all
the systems and all the phenomenaology, that have a
first order effect. And that's very straight forward.
And we want to also represent those which clearly have
a second order effect, which means that they impact
things in the order of 10 percent.
MEMBER KRESS: When you say systems, you
mean things like fan coolers and sprays?
MR. HENRY: Sprays, right. And, of
course, the M&E coming out of the break and any
special things. If we're looking at another plant,
like Cook, the dynamics of the ice condenser and its
melt and drainage, and etcetera.
And the things that relate to 10 percent
that could be issues. Things like water entrainment
may either influence things by order of a 100 percent
or the order of 10 percent. As we'll see later on,
both nodalization of water entrainment have a
significant influence on this.
That's our real focus, to make sure that
we cover all these phenomena, and when you get down
things which relate to one percent, it's kind of hard
to deduce what kind of influence they really have and
then take the jump to full scale containment
experiments and try to look for that particular
effect.
CHAIRMAN WALLIS: The first order is 100
percent, and you can't have very many of those, can
you?
MR. HENRY: Correct. Don't have too many.
That's where order of magnitude comes in.
CHAIRMAN WALLIS: You said first order was
30 percent or something, then you could have three of
them.
MR. HENRY: That looks like an argument I
need to delegate to somebody who is younger and
quicker, Graham.
Okay. First off, but what's the influence
of nodalization? Because that's one of the aspects
that you just heard that's important to these two
sites. They currently are licensed with single node
models and it gives them some difficulties when they
take the current design basis in the M&Es and apply it
to the model. So is it the limitation of the model or
is it a limitation of the design? Well, the only way
you can figure that out is to do something which has
more than one node and look at the influence of it.
So what we did to just determine the
influence of moving to a multi-node containment was
take MAAP and let MAAP produce the M&E. And so this
is not coming from a design basis M&E, but it's not
meant to say this the plant response. All we want to
look at here is what's the influence of single node
verses multi-node for a large break LOCA response and
for main steam line break response.
What we have then --
CHAIRMAN WALLIS: What you want to do with
CFD is you keep applying the nodes until it makes no
difference. Here you're showing there is a
difference, but you don't show -- you keep on going to
a 100 or 200 or 300 node --
MEMBER KRESS: I think there's a
difference of what we're calling nodes. These nodes
have specific boundaries --
CHAIRMAN WALLIS: A physical basis.
MEMBER KRESS: Yes, physical basis. Those
ones you're talking about are kind of different.
There's a difference, I think.
MR. HENRY: And you're also taking away
any information I could use if you guys say come talk
to us again. I mean, that was one of the things I
want to do next time was show you how we progressed.
In any way case, we're going to get to
that in a little bit. And unfortunately I didn't have
it put in here, but try to look at the differences in
example CVTR going from one to 4 nodes and then 6
nodes. Six nodes which are this way and 6 nodes which
-- might be 2 this way and 4 this way and etcetera.
What you really find is you're not very sensitive to
that. What you're sensitive to is getting past one
node so you can have air move throughout the
containment.
If you have various rooms, then it's
certainly to your benefit to make those nodes, because
things could potentially be more concentrated in that
room if there's not sufficient natural circulation.
CHAIRMAN WALLIS: You might say it's
rather ridiculous if you take several rooms and mix up
all the atmospheres and then saying that that's
typical of everything that's going on. That's s very
crude and probably inappropriate way to look at what's
happening.
MEMBER KRESS: So it would depend on where
your break -- which room your break occurs in?
MR. HENRY: Slightly. And as Tom said --
Mike said, excuse me. Your Tom, right?
CHAIRMAN WALLIS: But the room where the
break is very different from the rest of the
containment?
MR. HENRY: I think Tom's point is we look
at a break in each of the three different compartments
for Beaver Valley, as an example. And that's part of
the analysis. But it's not greatly different between
them, but it is tenths of psi difference because you
don't necessarily get the same condensing profile
throughout the containment depending upon where the
break is. Because even though the compartments you
might think are equivalent, but they don't necessarily
have the same entry area and existing area, etcetera.
Anyway, to the point of nodalization, this
is a demonstrative calculation. This happens to be a
Westinghouse two LOOP plant, and we divided this up
into 5 nodes and also ran it with 1 node. So one of
the nodes is, of course, the reactor cavity. The
second node is the loop compartment which houses the
two loops. The third node is the annular region which
is outside the loop compartment. And then the
operating deck, which is here, we put two nodes in.
one above the operating deck, one here and one in the
region above the spring line.
CHAIRMAN WALLIS: I have to ask why would
you ever mix 1 and 5 in any kind of node?
MR. HENRY: I think the answer to that is
originally when people did design basis calculations
that was judged to be conservative, give you a higher
peak pressure. And from a practical point of view it
certainly makes sense that you would always have the
same air pressure, pressure everywhere, so it limits
the condensation rates.
CHAIRMAN WALLIS: So the big action role
of this is the condensation on structures and that
sort of thing?
MR. HENRY: Yes, the big actor is
condensation on heat sinks. That's what really drives
the bus on whether or not you live within your current
design basis pressure differential. And some of that
is shown in this slide. What we have here is really
single node and multi-node, which is shown here when
it says 5 node and 1 node.
So if I take these two, which says 5 node
and 1 node, which is this solid line and this large
dashed line here. This is MAAP4, and I apologize I
didn't get that written on there, but that does not
have things related to atmospheric circulation to
water entraining. And as we walk through this you'll
see some of those influences. Whereas there here are
MAAP5, which is the design basis code that we're
looking for for these two sites. And you'll notice
that you can identify, this induced flow is equal to
1 means the induced flow from the break was included
here.
But as you look at this for 5 node with
MAAP4, this solid line, and 5 node with MAAP5 you see
no difference. And it's true. Because the only real
thing that made a difference here was going from 1
node to 5 node reduced the peak pressure
substantially, roughly in atmosphere. And the whole
reason is that in the local near the break you pushed
air away and you got enhanced condensation during this
short time frame of about 10 seconds. And that makes
a difference.
And all we're doing here, it's the same
code, it's the same physics. Obviously, we're just
changing the number of nodes. So you can see even by
including all these new models we're going to talk
about from MAAP5 that didn't make any difference and
it's strictly the single node going from 1 node to
multi-node that made the difference.
CHAIRMAN WALLIS: This is pressure
absolute?
MR. HENRY: Correct. This is pressure
absolute here.
CHAIRMAN WALLIS: But it says G on the
other side, right?
MR. HENRY: And here, as you can see, this
is one atmosphere there. This is .5 and 1.5 times 10
to the fifth. This is absolute and SI units and we've
put it in gauge over here.
For this particular plant, the design
basis pressure is 60. But, again, that's for --
CHAIRMAN WALLIS: 60 psig?
MR. HENRY: 60 psig. And that's for
design basis mass and energy increases, which are not
in this calculation. That was not the intent here.
The only intent was to illustrate the difference of
going to multi-node.
And, as you might expect, for a large
break LOCA this is just the temperatures in
containment, again, in terms of Kelvin and Fahrenheit.
There's really not much difference between the two
codes.
CHAIRMAN WALLIS: Of which node?
MR. HENRY: Excuse me, Graham.
CHAIRMAN WALLIS: Which node? Temperature
in the containment is different in different nodes.
MR. HENRY: Well, of course, this only has
one node. This is the lower compartment, so this
would be node number--
CHAIRMAN WALLIS: It's different nodes.
Okay.
MR. HENRY: Right. This is 2. I think
that's 4 and 3.
But in essence it says that there's not a
big difference between them, and that's not surprising
for a large break LOCA, because the blowdown itself
puts so much moisture into the atmosphere.
So then we take the same analysis, again,
just from a demonstrative point of view what does it
mean for main steam line breaks, and that's a little
different story then. But here we have these two are
MAAP4, that have nothing after them. And these that
say induced flow=1, this is MAAP5, which again is the
code that we're talking about here.
CHAIRMAN WALLIS: I only see 3 curves.
MR. HENRY: You always take my punchline,
Graham. This 1 node curve and this 1 node curve are
on top of each other. It doesn't make any difference
from a practical standpoint. And the reason this is
different here is now, as we'll get to later on,
what's influential in MAAP5 is induced circulation
because a main steam line break goes on for a lot
longer time. But if you say there's only 1 node
available, then as a result of that you're doing this
momentum equation into one huge node of the
containment, and really it's a huge mass and it hardly
stirs it all all and issues related to enhancing any
local velocities or entrainment really go away. So
that part really disappears in one node and they
become the same calculation.
But when you go to 5 nodes now, of course,
the blowdown is coming into one of those nodes, which
is a much smaller region and also you, obviously, have
higher heat transfer in that local. Because it is 5
nodes you're displacing air away from it and you have
the potential for also reentraining moisture in the
containment because a local velocity in that node can
be above an entrainment criteria.
MEMBER KRESS: Does that act like a water
spray?
MR. HENRY: Exactly.
MEMBER KRESS: Does the code then account
for revaporization of the droplets?
MR. HENRY: They can allow the droplets to
revaporization. But principally when you entrain
something, you're entraining the film off the wall, so
you entrain at the average temperature, which is T
side on the outside and T wall. So you actually get
some subcooling.
MEMBER KRESS: So you get some subcooling.
MR. HENRY: Which really is the major
thing to do. But yes, Tom, it can revaporize. In
fact, that's part of what you see with multi-node
because it can get down to something in the bottom of
the containment for local partial pressure is not so
high and the droplet might be warmer and it can
vaporize down there as it falls through that node.
MEMBER KRESS: Do you have a model for the
rate of entrainment when the droplet --
MR. HENRY: At the rate of entrainment and
we put that into -- it goes directly into the aerosol
model where the deposition rate is depending upon the
airborne density. So the airborne density --
MEMBER KRESS: So you exercise -- is this
still the aerosol model that was built by --
MR. HENRY: Mike Epstein. Yes. So it
becomes -- water is just part of the aerosol --
MEMBER KRESS: Part of the aerosol?
MR. HENRY: Right. But the only reason I
wanted to make a point, is the aerosol can come from
either entrainment or from cooling of steam, both of
them get put into the aerosol mix.
And this, Graham, to go back again, we
have 4. The piece in the legend, the two curves are
simultaneous on top of each other because here again,
the circulation has no influence. It's so slow. And
what we get out of that since it's also not
entraining, you get temperatures which are typical of
what you see in some of the main steam line breaks.
CHAIRMAN WALLIS: I don't know how we
would apply it, the circulation model to those five
different rooms and there's not going to be one big
circulation pattern to these five rooms. It just does
not plot.
MR. HENRY: I agree with that. Just from
the concept if I assume that it applies, it says it's
not going to make any difference anyway because I'm
too big, you can't make me circulate fast enough. But
if it does and it entrains, you can see what's gained
on the peak temperature, so it's again substantial. So
not only is the pressure lower than the 1 node system,
the temperature is also lower.
CHAIRMAN WALLIS: These are all, of
course, predictions?
MR. HENRY: These are calculations, right.
We're going to get to comparing this with experiments.
I just hesitate saying predictions because
it's really for a generitized system. It's a two LOOP
plant, but it's different from Point Beach's
containment model in terms of nodes and level of
qualification.
And also then we just look at for main
steam line break between MAAP4 for a 5 node model and
MAAP5 for 5 node model. And this all becomes because
MAAP4 knows nothing about atmospheric circulation,
knows nothing about entrainment model. And all we're
comparing here is the influence temperature of those
particular models which is what we'll talk about
today. And this has the multi-node in it, but it
still isn't enough to really -- I'll get to CVTR. If
I take this approach from MAAP5, which we thought was
a quite good code when we started that comparison, it
overstates the pressure in CVTR by something in the
range of 10 psi and it overstates the temperature by--
I forget the actual number. Like 50/60 degrees
Fahrenheit. This is the physics that we believe is
controlling that.
So I mentioned MAAP is not a 1 mode model,
so this isn't meant to -- this just shows you the
various pieces of physics that are in the model, and
some of these are severe accident related which Tom
correctly asked us earlier what's being reviewed. And
what's being reviewed is that the containment model as
it applies to design basis accidents.
Now, the key actors in that for these two
different plants, we obviously focused on the heat
transfer to heat sinks, which is shown here. Tom, the
aerosol model we just talked about, which is part of
this.
The heat transfer to equipment, which is
just all the steel and everything inside, whether it's
handrails or ducting or whatever it may be.
Condensation on all the walls and on all
the heat sinks. The metal, concrete, steel lined
concrete, stainless steel line for fueling pools.
Fan coolers for Point Beach. And, of
course, the sprays for both Beaver Valley and Point
Beach.
And lastly, the flow from the primary
system. This is not coming from MAAP now when we talk
about design basis things. This is coming from
Westinghouse design basis mass and energy release
calculation for both large break LOCA and the main
steam line break.
MEMBER KRESS: This allows the use of
sprays and fan coolers. Is there a single criteria in
the DBA that says you can't use the full capacity on
those?
MR. HENRY: Correct.
MEMBER KRESS: That's probably --
MR. HENRY: There's a whole run matrix
that's used by both of the sites depending on what
their specific conditions are. They look at all the
different kinds of single failures and look for the
worst one in both sets of conditions.
And then also, some temperature is part of
that, so that has another set of M&Es or way that you
treat the previous M&Es, to mix or not mix them coming
out of here.
And, again, we talked about uncertainties
in the models, but there's also variations in the
operating perimeters that have to be part of that DBA
calculation. You have to look for the most limiting
case of operating conditions.
MEMBER KRESS: I guess I would look for a
net positive suction head located in those
compartments that could prove affected.
MR. HENRY: Well, the analyses that you
look for is the net positive suction head when you go
into recirc there for sure, yes.
MEMBER KRESS: But you look at that?
MR. HENRY: Yes, that's part of the --
MEMBER KRESS: You don't --
MR. HENRY: Yes. And you look at it on a
plant specific basis. Because even when you have two
units at the same site, they don't necessarily have
the same systems.
Okay. So this is the conceptual part then
of the circulation, which is one of the key things
that we think is missing in MAAP4 and it wouldn't buy
you anything if you just looked at a one node anyway.
But the concept that it has is that a blowdown into
this gaseous region adds momentum to the atmosphere.
Obviously, if we just had one node and we have
momentum going in, where did it go?
CHAIRMAN WALLIS: These are different
nodes in the sense that there's previous nodes or are
these different nodes within a given compartment?
MR. HENRY: These are nodes in the same
sense as the previous nodes.
CHAIRMAN WALLIS: So there are rooms?
These are four different rooms?
MR. HENRY: They may be rooms or they may
-- this node boundary may be drawn in the atmosphere,
as an example.
CHAIRMAN WALLIS: I find this an
extraordinary diagram. I mean, the idea that there
are rotating cylinders in each one of these rooms.
Fantastic. And the idea that the incoming flow coming
up like that rotates the cylinder on top of it is also
fantastic. And the idea that nothing happens between
them except interfacial shear is also fantastic.
I couldn't understand what you could
possibly be showing. This is sort of a study of what
it sees in liquid helium or something.
MEMBER KRESS: You just conserve momentum.
CHAIRMAN WALLIS: No, there's no momentum.
It's on the angular momentum.
MEMBER KRESS: You've got momentum coming
in to flow and you're going to put that all in the
atmosphere until it circulates.
CHAIRMAN WALLIS: It doesn't happen that
way, it's all angular.
MR. HENRY: Well the angular momentum is
still momentum. Graham, this is merely now to
describe the concept. Because the concept is it will
-- let's first just think of a single node here as an
example.
CHAIRMAN WALLIS: But that's not the way
it works.
MR. HENRY: I know it isn't. I know it
isn't. But from a single node point of view, it's
easier to see what happened to the momentum that came
in here. Where did it all go? Because in a single
node you now have inflow, where did all that momentum
go? We can have conservation mass and conservation of
energy, where did the momentum go? It has to go in
terms of somehow this fluid is circulating.
CHAIRMAN WALLIS: I thought you were going
to say, the incoming flow in region 2 there actually
set up some sort of a circulation around the jet which
helped the heat transfer to the wall?
MR. HENRY: It certainly does that. This
doesn't mean that this sits here and spins with either
a sphere or cylinder, whichever you choose on that
one. It only means -- it only gets down to this
fundamental thing right here. Schematically what the
code thinks of is I've got some velocity in this node
which is different from the through flow velocity. I
have circulation. And this is merely a way of
representing that, but that momentum that gets added
to that node says the only way I can satisfy my
momentum balance is I've got to circulate faster.
CHAIRMAN WALLIS: I don't understand that.
MEMBER KRESS: Well, it bothered me when
you said that those nodes could be virtual nodes in
the middle of the air and still do that. Those are
not really boundaries of rooms?
MR. HENRY: Certainly whenever it is a
boundary of a room, you use the boundary of the room.
But sooner or later you'll have to draw if you want to
be able to investigate whether or not things can be
stratified.
MEMBER KRESS: I see.
MR. HENRY: You have to eventually draw
something up here, which is air. Otherwise you're
always just going to have rooms and this will always
be one node and you won't have any stratification
potentially.
MEMBER KRESS: But you can't treat that in
terms of momentum the same way --
MR. HENRY: Let's talk about it.
CHAIRMAN WALLIS: Let's go back to this
other picture that you showed us. I don't understand
it. You've got nodes and you've got flow between
nodes, which is the usual thing.
MR. HENRY: And that's part of the
calculation.
CHAIRMAN WALLIS: And within each node you
have some sort of circulation as well? Is that the
idea?
MR. HENRY: Yes. Because if you have flow
coming in and you conserve mass and energy, so what
you have going out of here is merely the through flow,
then you won't be conserving momentum.
CHAIRMAN WALLIS: Well, that's because it
forces on the wall.
MR. HENRY: No, even without that. You're
just going to defuse or any momentum goes away. You
don't satisfy it by the through flow alone.
CHAIRMAN WALLIS: You're really confusing
me altogether.
MR. HENRY: Okay. That's tough to do.
You're a hard guy to confuse.
CHAIRMAN WALLIS: No. I mean the momentum
balance works out always. If you don't have -- if the
momentum balance doesn't work out you've got forces of
some sort. The idea that the linear momentum is
balanced by angular momentum is a very strange
concept. So something else is going on.
I think what you're saying is that the
incoming flow in to that chamber stirs things up so
the fact that this sort of -- some average velocity,
which is low, is not characteristic of the real
velocity seen by the wall. Isn't that what you're
saying?
MR. HENRY: Well, in a sense yes. That's
why I wanted to go back to just from a simple concept
look at a single node. If I did an experiment with a
single node. And I blew down into that. What would
be the governing velocity of the through flow that
comes out, if I make it one dimensional. Of course
not.
MEMBER SCHROCK: The way you've run it, it
doesn't look like there should be any shear between
nodes in that picture.
MR. HENRY: Well, we're going to get to
that.
MEMBER SCHROCK: You're both going in the
same direction.
MR. HENRY: Right. Yes, they're going in
the same direction, but they don't necessarily have to
be going at the velocity.
Suppose I put a bunch of structure up in
this node, as an example, Virgil. So this may be
going at a much higher velocity than this. I still
have to represent the fact that there could be
momentum transfer across this arbitrary boundary that
the nodalization has created. That's all it really
means.
MEMBER SCHROCK: How you come by those
velocities, you're going to show us.
MR. HENRY: Sure. I knew you'd guys would
have tons of questions on this, and that's why we're
here.
CHAIRMAN WALLIS: Well, I just understood,
let's go to the next equation.
MR. HENRY: Okay. I would just say,
before we leave this, Graham, you said it very well.
All this is meant to merely say -- all this says is
that a node has a property that we looked at as
circulation. And that's merely the way of making sure
that we do conserve momentum throughout these various
nodes.
CHAIRMAN WALLIS: Circulation cannot
conserve momentum.
MR. HENRY: Okay.
CHAIRMAN WALLIS: Circulation cannot
conserve linear momentum.
MR. HENRY: Yes, you're right. It does
not conserve linear momentum, but --
CHAIRMAN WALLIS: Spin those things up to
the speed of light, and it won't conserve --
MR. HENRY: That's why I wanted to make
sure we talked about the single node. Within a single
node is an example there can't be any linear momentum.
There's no out flow. The only thing you could have is
something that goes back to that's going to spin it
somehow or other. But you know from those kinds of
experiments that you have a lot higher energy transfer
at the wall, and that you get by the through flow
velocity or the pressurization velocity.
CHAIRMAN WALLIS: What is this -- can we
go to the first line here?
MR. HENRY: Sure.
CHAIRMAN WALLIS: What's going on here?
MR. HENRY: All this does is say that the
way we look at this is the equation -- equating the
impulse and the rate of change in that specific node.
What's it's mass and what it's velocity.
CHAIRMAN WALLIS: "U" is a circulation
velocity?
MR. HENRY: Yes.
CHAIRMAN WALLIS: So this is spinning?
MR. HENRY: It's a concept of there's
something going on and whether it's one thing or
spinning this way or whatever, it's not a through flow
velocity.
MEMBER KRESS: It's not spinning. It's
just falling circle. That's different than spinning.
MR. HENRY: Okay.
MEMBER KRESS: It's not angular momentum
--
CHAIRMAN WALLIS: If you give something an
impulse, it moves linearly, it doesn't --
MR. HENRY: Yes, it does.
MEMBER KRESS: Well, this is a linear
motion. But I don't know what "F" is yet, that's
what's bothering me.
MR. HENRY: We have three different forces
that we look at, which is the force on the wall here,
on the shear force on the adjacent node, which Virgil
was asking about. So if you have a difference in the
rate at which you have the circulation velocity and
nodes, then that has its own influence. And then if
you could have any kind of embedded structures that
slow things down, they also have to be --
CHAIRMAN WALLIS: What is Uc in your
figure?
MR. HENRY: Uc is the property in the node
which is --
MEMBER KRESS: That's the result you're
trying to calculate, right?
CHAIRMAN WALLIS: What is Uc in this
figure?
MR. HENRY: Uc in the concept of the model
is that in addition to through flow that this is also
has --
CHAIRMAN WALLIS: Where is the Uc? I mean
is it an average of some sort or is it on the wall, or
in the middle. Where is Uc? I don't understand.
MR. HENRY: Okay. Uc when we look at the
energy transfer to this wall right here, Uc is the
velocity that's dictating what the boundary line --
CHAIRMAN WALLIS: Uc is the velocity along
the wall?
MR. HENRY: Uc is the free stream velocity
next to the wall.
MEMBER KRESS: It has to be some integral
of the velocity in the whole mass --
MR. HENRY: It is, yes, right, Tom. And
that's because it's coming from a momentum balance on
each node.
When you have a through flow velocity and
in each node you have a property called circulation.
And whatever that velocity is, that's what determines
the free stream velocity next to the wall, it also
determines the velocity that could entrain anything
that's collected in that node. Reentrain water, which
is what --
CHAIRMAN WALLIS: What you would call a
turbulence velocity or be about the same thing?
MR. HENRY: Yes.
CHAIRMAN WALLIS: It would be about the
same thing?
MR. HENRY: Exactly the same thing.
CHAIRMAN WALLIS: It is the amount of
stirring up of the nodes, a measure of the stirring up
of the fluid in the node by incoming flow?
MR. HENRY: Exactly. Exactly. The only
reason I pictured it this way is to try to break it
down to the most simple thing. The code thinks I have
a velocity here and so where does that go. Well, it's
evaluating as if it is stirring or a turbulence
velocity. It's not the through flow velocity to the
next node.
CHAIRMAN WALLIS: Well, I think you're
going to have to look at the details of this somehow,
because, you know -- it may be a brilliant idea, but
I'm having trouble understanding it especially treated
like this. There's no way that incoming flow going
straight up there is going to stop, swirling things
around in the way you've drawn that.
If you had said there was a level of
turbulence, a mixing or something, I think I might
come closer to understanding what you mean.
MR. HENRY: But that's exactly what it is.
But the code has to have some concept that you're
loyal to and how it incorporates this information into
-- the information flow of what you're actually
calculating.
MEMBER KRESS: Did you mean for your Ms in
that first equation -- second equation to be under the
parentheses?
MR. HENRY: This is what the M is, this
has the same units of force. This is kilograms per
second and --
CHAIRMAN WALLIS: There's no way that in
the way you've described Uc that that first equation
you've got comes from a control volume analysis. It
comes from some kind of a word picture of some kind.
There's no way you can draw those Fs on a box and show
me how the linear force produces angular momentum.
MR. HENRY: Graham, I completely agree
with that. You won't be able to take this into
something and say, gee look that's now angular
momentum. But by the same token, when it -- when you
hit all these structures, and I'm just trying to
follow through what you've done, is you've created
turbulence. So some way this thing has a velocity
that's different than the through flow velocity.
CHAIRMAN WALLIS: I think that this is
important and you're going to have to establish there
some sort of a believable, mechanical basis for these
Ucs in terms of physical phenomena.
MR. HENRY: Absolutely.
CHAIRMAN WALLIS: Because I'm not looking
for something that's academic and terribly fancy --
MR. HENRY: I know.
CHAIRMAN WALLIS: But this seems to be
fanciful.
MR. HENRY: We'll take that as an action
item. When you see us the next time we'll go through
how we get to that.
CHAIRMAN WALLIS: We have to sort of buy
off on this. It may turn out to be a brilliant move
in terms of a way out of the box, so you're going to
have to represent something which is important
physically.
MR. HENRY: That's a very good way of
putting it. You have to find something that's
consistent with this big thing that represents 17
nodes. What's going on. Right. What's going on.
MEMBER KRESS: And I think this is a
reasonable concept if you have real boundaries. But
I'm not sure when you stick these virtual boundaries
in --
MEMBER SCHROCK: Yes, I'm having the same
problem.
MEMBER KRESS: That's why I think I need
to see this, the validation.
CHAIRMAN WALLIS: I'd like to see it also
--
MEMBER KRESS: I think it can be done,
really.
MR. HENRY: Well, we have some experiments
here that focus on just that thing.
CHAIRMAN WALLIS: I think what you're
saying is if you open this door and you open that door
and there's a draft going through here, it stirs up
the fluid in the corners as well. Is that the sort of
thing you're saying?
MR. HENRY: Yes. And the rate at which it
stirs it up is dependent upon the -- you can't get it
from mass balance.
MEMBER SHACK: But you're saying that the
stirring is related to the momentum? You don't get
the stirring without some momentum.
MEMBER KRESS: He actually has another
equation that calculates this momentum going out.
He's got momentum coming in and going out. It's the
difference between those that goes into the stirring
up. It is sort of an integral -- it's an integral
amount. It has to go somewhere.
CHAIRMAN WALLIS: Because you always get
the forces on the wall.
MEMBER KRESS: I'm ignoring those.
CHAIRMAN WALLIS: You ought to give him a
D for that.
MR. HENRY: I knew we were going to have
a lot of questions on this. You guys are true to
form. You're still younger and quicker than I am.
Graham, what actually happens here if I
take all of this out of here, is I would expect this
to be a jet which begins to entrain as it goes up
through here. It entrains on the way up and it hits
up here, and it spreads and it comes down. But all
that ends up being, stirring of this atmosphere, and
stirring eventually -- well, basically it hits this
wall and you take momentum out of it and you start
turning it angular momentum now.
The way the code has a concept of that,
because you can't -- it's very difficult to put all
this kind of structural detail in --
CHAIRMAN WALLIS: The code doesn't have
any concept. You write the concepts.
MR. HENRY: Okay. You're right. The way
my code has a concept --
CHAIRMAN WALLIS: No. The way you imagine
it does. Let's get it clear: This is some kind of a
Henry fantasy.
MR. HENRY: This goes way back. We've had
a lot of fun with this over the years. So it's always
been his fun, though.
Anyway, what is imagined for this then and
the way it gets incorporated into the code is instead
of trying to represent all this through detail, as an
example, for jet flow etcetera, is to put this in
something that says okay let's do the momentum balance
on this and it will be interpreted as a velocity,
which is turbulence, circulation and that velocity,
that influence is what's used to determine the shear
on the wall, the energy transfer to the wall and also
it's ability to entrain.
So, that's why I put this up as a concept
trying to put this into a large code that you could
easily track through what it is or what its influence
is and what are all the things related to slowing it
down, whether the influence is out of entrained
structures. But you eventually have to get to drawing
boundaries in the air someplace or you won't have
stratification.
CHAIRMAN WALLIS: It would be easier for
me if instead of calling it circulation you said
there's a schematic of -- there's a mixing, mixing
velocities which are produced by the flows or
something like that. The idea of the circulation with
these big cylinders rotating is something that I have
trouble with. But if you said -- same as you got flow
in a pipe where the transfer to the wall it's governed
by the turbulence which it's sort of set up by the
main flow and you just don't say it's a linear flow
because then you wouldn't have transfer to the wall at
all. Let us somehow model the turbulence. I think
that's what you're trying to do.
MR. HENRY: That is what we're trying to
do.
MEMBER KRESS: But tell me, how do you get
the momentum out?
MR. HENRY: Well, see, I've got to get
even with him next time. We have a slide that's
nothing but words and he's going to say can't you draw
me a simple picture of this.
Okay, Tom.
MEMBER KRESS: How do you get the momentum
out there with a pressure difference in an area, a
lost coefficient bobbing between the node?
MR. HENRY: Yes.
MEMBER KRESS: So really --
MR. HENRY: Since it is a pressure
different --
MEMBER KRESS: You don't have any screening
in the momentum code.
MR. HENRY: We have the pressure
difference that says what is the flow rate that's
leaving the node.
MEMBER KRESS: Okay.
MR. HENRY: And that's evaluated. And
what it carries with it is whatever that turbulence
velocity is.
MEMBER KRESS: It carries its node
velocity with it?
MR. HENRY: Yes.
MEMBER KRESS: Computing the pressure
difference?
MR. HENRY: The pressure difference --
MEMBER KRESS: Is this that lost
coefficient?
MR. HENRY: Yes. Yes, there's a lost
coefficient if it's just wide open, then there's
basically no lost coefficient. But, you know, the
real fundamental thing at least we've discovered here,
and that's what I want to also verify to myself as we
work through this is what was the insight here was if
you didn't this, then you never got the right answer.
If you did have it, it didn't make much difference how
much detail you went to as long as you said someplace
that momentum got observed and therefore we had
turbulence velocity which was higher than just the
through flow velocity.
MEMBER LEITCH: You've talked about this
containment of pressure and temperature in macroscopic
sense but then do you calculate pressure and
temperature in each one of these virtual nodes, or
which one --
MR. HENRY: Each node has its own
pressure.
MEMBER LEITCH: And I guess my question
then is that it seems to me to say in the LOCA, that's
where the LOCA occurs, you would have a higher
pressure and temperature.
MR. HENRY: It does.
MEMBER LEITCH: Does that become limiting?
MR. HENRY: Generally not because it's
usually the saturation temperature corresponding to
the pressure in the room, and that's what we figure
with most plant's design basis already is. But
certainly the pressure and the temperature in the
break room are highest. We'll get to a little bit of
that later on.
MEMBER LEITCH: Higher than the previous
methodology that are indicated?
MR. HENRY: Lower. Lower pressure and
some are lower temperature than previous
methodologies. Because you get more condensation by
displacing the air. Temperature is also mitigated
because of all the moisture that gets entrained back
into the atmosphere. So it's hard to ever have super
heat, which is again what the experiments seek.
MEMBER KRESS: When you say dry runs is
submerged pressure, what does this submerged mean
here?
MR. HENRY: It could be things like
grading, I-beams.
MEMBER KRESS: Submerged means it's just
in there --
MR. HENRY: This is submerged in the air
right here, as an example.
MEMBER KRESS: Okay. It didn't mean it
was under liquid?
MR. HENRY: No. No, it's just submerged
in whatever the local fluid is.
MEMBER KRESS: Submerged surfaces normally
is a function of exposed surface area.
MR. HENRY: Right.
MEMBER KRESS: It depends on which
direction the flow is going. Does the code recognize
flow direction somehow and --
MR. HENRY: No. No, it just thinks it's--
MEMBER KRESS: It takes the polarity of
whatever the structure is --
MR. HENRY: If you have a pipe that runs
through the room, you know, it doesn't care whether
it's horizontal or vertical it has this turbulence
velocity that's used. We value how fast the it can
slow itself down.
CHAIRMAN WALLIS: How do you get drag on
submerged structures? It's the circulation velocity
that's dragging on this structure or --
MR. HENRY: Yes.
MEMBER KRESS: Do you use some sort of
friction lost coefficient or --
MR. HENRY: Just drag coefficient.
MEMBER KRESS: Do you have form losses in
it?
MR. HENRY: Well, it take it -- basically,
again, it comes down to if you put it in the code,
once you have it in it doesn't matter a whole lot on
details. But what we really use are just drag
coefficients associated whether we think it's a
cylinder or a square, or -- it's usually a pipe or
some kind of I-beam or grading --
MEMBER KRESS: So there is some sort of
consideration of flow direction versus the orientation
MR. HENRY: Yes. Again, yes. I did not
answer your first question right. We're always
assuming it's going across it, it's not going with.
You asked me a question, I responded incorrectly.
CHAIRMAN WALLIS: I think you have
something like a K epsilon here. You're saying that
the turbulence level in these nodes is a source of
energy to be fed in to increase the turbulence which
is the flows and then there's various frictions and so
on are dissipating turbulence. So you get some
measure then of atypical mixing velocity within the
node. I think that's the kind thing you're doing
here?
MR. HENRY: Yes. As opposed to saying
it's only the through flow velocity, which I'll come
back to in a second.
One other aspect is the condensation
occurs under natural convection conditions. In MAAP
we use the analogy for between heat to mass transfer.
So the thermal boundary layer is the same as what we
have for the natural circulation flow. Of course,
under laminar conditions, the Nussel number for all
gaseous flow -- excuse me, for single phase flow is
proportion to one-fourth power and then turbulent flow
we have the lower Reliegh numbers, the one-third power
at the higher Reliegh number, about .4 power which
comes out of standard textbooks.
And what we use for that, there's the
maximum of all these, depending upon what your
specific conditions are. The Reliegh number --
CHAIRMAN WALLIS: Your whole idea of
having circulation velocity is that the stirring
enhances the forced convection and produces the
transfer to the wall.
MR. HENRY: Right.
CHAIRMAN WALLIS: Now you're bringing in
pre-convection --
MR. HENRY: There are times where stirring
velocity is so slow it has no real relevance. It
eventually dies away, in other words. But there are
times when this is the governing process of energy
transfer to the wall.
CHAIRMAN WALLIS: So you have mixed
natural convection and stirring convection --
velocity?
MR. HENRY: You could also put here --
CHAIRMAN WALLIS: Is this why you put the
max in here, is that --
MR. HENRY: Yes. Because as the velocity
dies away, then this natural convection will take
over. So you have to have a consistent way of
addressing that as well.
CHAIRMAN WALLIS: There could be a
condition where the circulation would actually act in
the opposite direction of the natural convection and--
MR. HENRY: Right.
CHAIRMAN WALLIS: -- the net result would
be to reduce the heat transfer.
MR. HENRY: Right. That's one of these
pieces right here. When we use this, as I'll show you
in a second, which is just a straightforward saying
this looks nothing more -- it doesn't know that
there's a film on the wall, you just have natural
convection driven by the temperature difference and,
therefore, what's the hydrodynamic boundary layer,
what's the mass transfer boundary layer. We find that
we under predict the condensation rate when we go to
specific experiments, separate effects experiments.
MEMBER KRESS: What do you use for L?
MR. HENRY: This is the height of the
wall.
MEMBER KRESS: The height of the node if
it's the virtual node.
MR. HENRY: Yes. It almost cancels itself
out, as you know.
So what we have -- we'll get to, is it's
strictly correlating factor that says okay what are
our differences. And when we look at the data, the
higher the mole fraction of steam, the worse we do in
this straightforward representation of going from
single phase -- the heat transfer analogy, applying
this single phase gaseous representation to the
condensing potential.
MEMBER SCHROCK: What do you do for
condensation on your horizonal surfaces?
MR. HENRY: The condensation on horizonal
surfaces, it's usually dictated by the conduction on
the surface. That's water, it's very low, of course.
And if it's vertically -- if we're on a ceiling which
is facing downward, then we end up using this same
thing for experiments that I'll get to later.
Because, obviously, we have ceilings to the
containment and --
MEMBER SCHROCK: These natural convection
for mass transfer formulations are not appropriate
then?
MR. HENRY: That's right. And that's why
we go to the experiments. When we get to these
ceiling, which are facing downward, that is the
representation that they see in those particular
experiments.
CHAIRMAN WALLIS: I guess your document
explains this Fm so I can understand it?
MR. HENRY: This Fm is right here. This
is the correlating parameter, this is merely a
viscosity radiogram that says --
CHAIRMAN WALLIS: Oh, that's viscosity?
MR. HENRY: This is a viscosity ration,
this is to say this Nu of the gas over Nu of the
fluid. It merely gives us -- this is the most -- as
we'll see in a second, this is the most effected
parameter here that says the more steam you have, the
worse this representation does.
CHAIRMAN WALLIS: What is N?
MR. HENRY: That is the mole fraction of
steam.
CHAIRMAN WALLIS: So there's something in
your documentation that justifies this equation
somewhere?
MR. HENRY: This is strictly --
CHAIRMAN WALLIS: So it explains where it
came from in your documentation? If we've got the
code documentation, could we understand where that
came from?
MR. HENRY: I hope so. This is really
just a correlation for -- this is dimensionless
obviously, and these which we -- the viscosity ratio
because we have to cover all pressure levels here, the
reason this is to the .8 is viscosity squared to the
.4 power. And this is linear because all this is is
the fact that if you have low density gases that are
being condensed, they can collect in the boundary
layer and they can impose the natural convection which
is going on. And there's a ton of papers in
literature that say this virtually cancels itself out,
and it does.
MEMBER KRESS: What happens to things that
condense on the ceiling and other horizontal surfaces?
MR. HENRY: Let me come back to that when
we get to the experiments in a second, if you would.
Because we're going to certainly come back to that.
I just wanted you to understand when we
get to natural convection, there is an enhancement to
the condensation rate to the natural convection side
that, again, comes from separate effects.
In the interest of time, I won't spend a
lot of time on this because we already talked about
it. The mass energy releases comes from design basis
calculations as they are applied to the containment
models for both sites.
And I'm not going to spend a lot of time
on this one either, because it really says much the
same thing. We look at all these types of accidents
and as a result we'd like to find all the experiments
that we could find that are applicable to these kind
to test the total capabilities of the containment
model.
The fact that the design bases mass energy
releases come from separate models and they get their
input to the containment model. So in essence what
MAAP is calculating for the core in the RCS is just
thrown away. It's ignored.
There's a mass energy release time
dependent mass energy that's coming into the
containment. That's exactly what we do to benchmark
the calculation against these major experiments of
CVTR and HDR. We have the mass energy releases which
are specified by the experimenters.
So I think I came to you guys once before
and said would it be worthwhile to try to make sure
that we preserve some of this very key data and put it
in the codes and that's really what we're trying to do
here as well. So it's preservation activity as well
as a convenient way to benchmark the code on a
continual basis.
Experiments that we currently have pulled
together, and this again, as I said, this is one of
the key places that we want to have your feedback, is
separate effects.
We've used the Dehbi condensation
experiments at MIT, the Anderson condensation
experiments at Wisconsin, the Hitachi condensation
experiments which related to a containment test but it
gave us another separate set of tests that we could
compare the condensation model against under natural
convection. Uchida condensation experiments, Tagami.
When we get to the spray experiments they just lightly
touch on the nice thesis that was done in Canada by
Kulic for both single droplet as well as spray header
behavior. And for countercurrent natural circulation
we used the salt water, the brine water tests done by
Epstein and Kenton for countercurrent natural
circulation where you have both heavy over light as
well as heavy over light with a through flow induced
as well to assess that set of conditions that could
flood the natural circulation flow too.
So these are the separate effects tests
that we've built up to date.
The large scale integral tests include
small break, large break and main steam line like
conditions for HDR. CVTR tests -- I should say the
HDR tests are all international standard problems
also. CVTR tests are steam into a containment. Steam
came from an adjacent power plant.
And the containment standard problem tests
were done at the Battelle Frankfurt facility. There
are two different types of hooking up of that
particular set of containment compartments.
By doing these, of course, we're also
demonstrating the use of external M&Es, because that's
what these are.
MEMBER KRESS: Have you checked into the
Marveicken --
MR. HENRY: We have, and they're so
dominated by the suppression pool.
MEMBER KRESS: They are, yes.
MR. HENRY: But, indeed, those are ones
that we like to add to this whole thing but not so
much for these guys.
That's a good point, Tom, because I wanted
to -- there are a couple of other experiments here.
One in particular is a CSDF test at Hanford. While
it's ice condenser related, it certainly enables you
to see what's this code going to do for the natural
circulation flows that they put into those
compartments. So that's also part of it, but not
listed here.
And the separate effects, one is the
experiments that we used heavily were the experiments
performed by Dehbi at MIT. And this was interesting
to us because you had a very long condensing length,
even though this is maybe something like an inch to an
inch and a half or so, but it had 3.5 meter condensing
length that gave nice natural convection conditions to
benchmark the model against. And they also,
obviously, had air as the noncondensible gas and they
put in light gas to see what the influence was with
helium also.
MEMBER KRESS: The vials for the outer
chamber heated or insulated --
MR. HENRY: Yes. These were insulated
here so that the steam came from boiling water and the
cold water was flowing through this copper condensing
cylinder that they have here.
CHAIRMAN WALLIS: And you put in a certain
amount of air so you have some noncondensibles?
MR. HENRY: Right. And in some case they
have a set of experiments where they bled steam
through the boiling water as well.
CHAIRMAN WALLIS: Now your code with nodes
in it, now it really doesn't address the question of
how do you predict the heat transfer coefficient in a
geometry like this, does it?
MR. HENRY: Well, natural convection heat
transfer coefficient that I just showed you, you could
either benchmark it based upon the condensing
coefficient on the wall just due to the natural
circulation condition --
CHAIRMAN WALLIS: Are there correlations
for a cylinder inside a cylinder or this kind of
natural convection?
MR. HENRY: Well, I can show you exactly
what --
CHAIRMAN WALLIS: You borrowed them from
some other context, or something?
MR. HENRY: The size of the cylinder means
that this almost looks like a flat plate in terms of
what the -- vertical flat plate in terms of what the
natural convection is on the outside.
CHAIRMAN WALLIS: But it's long, so you
have flow --
MR. HENRY: You do the hand calculation,
this steaming rate is nowhere near what it takes to
flood the film. This steaming rate is very slow.
MEMBER SCHROCK: Does it count as
turbulent film?
MR. HENRY: The following film?
MEMBER SCHROCK: Yes.
MR. HENRY: Yes. The turbulent film gets
fit for the turbulence.
CHAIRMAN WALLIS: I didn't mean it that
way in terms of that sort of flooding. I mean you're
going to use some sort of Nussel numbers or something
or obtained from a correlation like the ones you've
showed us?
MR. HENRY: Yes. It comes directly from
those correlations.
CHAIRMAN WALLIS: Assuming that this is
the same as flat plate in an infinite environment?
MR. HENRY: Yes. That's the assumption.
And this is the data. I apologize, these
are pretty small figures, but this is at a pressure
4.5 atmospheres. In essence, one atmosphere of air.
This is at a pressure of 3 atmospheres and
1.5 atmospheres. For a variety of air mass fractions
this is the way the data was reported by the
experimenter.
And then this plus the two on the next
page have helium fractions, this being 1.7 and the
others the 4 something and 8 something percent helium.
Now, this solid line right here is MAAP4,
which is just those mass and energy, the analogy of
heat to mass transfer applied to this set of steam
conditions.
CHAIRMAN WALLIS: So what you're testing
is is this Fm?
MR. HENRY: In essence Fm comes from
these, Graham. That's the correlation that comes from
the separate effects tests. What's the reason that
this-- why don't these equations work, as an example.
Well, you can see, certainly, as we have more and more
steam in here, the difference between those equation
by themselves and the data increases.
CHAIRMAN WALLIS: When you say comparison
of MAAP with Dehbi's, the only thing that MAAP did was
introduce this Fm.
MR. HENRY: Right, and that's a
correlation that comes from this information.
CHAIRMAN WALLIS: The information itself
came from these data.
MR. HENRY: Right.
CHAIRMAN WALLIS: It ought to fit them, it
was itself derived from the data.
MR. HENRY: My only point here is to show
you this is all fit --
CHAIRMAN WALLIS: How well it does?
MR. HENRY: Yes, how well it does and the
fact that you've known correlations from day one,
right?
CHAIRMAN WALLIS: Yes.
MR. HENRY: This is a value of that Fm of
1. I'd like to get that as close to 1 as --
CHAIRMAN WALLIS: And what this does is it
justifies that Fm is a reasonable way of modeling
condensation.
MR. HENRY: Exactly. Exactly. It's
nothing to say this is how well this does. But then
I'm going to take this same thing to all the other
experiments before I ever apply it --
CHAIRMAN WALLIS: Okay, so now I begin to
understand. Because, you know, you send us our slides
ahead of time, which was a very good idea.
MR. HENRY: Obviously. Well, it's always
a good idea.
CHAIRMAN WALLIS: Well, I'll look at this
and I say what has this got to do with containment.
It's really a separate effects to get the condensation
coefficient.
MR. HENRY: Yes. And I should if I look
at this test, I should be able to go to other tests
and do just as well. If I don't do as well, I'd
better broaden these uncertainties --
CHAIRMAN WALLIS: So you're not modeling
any of the circulations or --
MR. HENRY: No, no, no. In fact, this is
really set up to be just natural convection is the
dominant thing.
This is a value of 1. This is a value of
that Fm of .5 and 1.5.
Well, certainly from the standpoint of
moving through various pressures, it does a reasonable
job of bounding the data so we can find out the role
of uncertainties or this uncertainty, where this
particular thing applies in a containment analysis.
But before we do that, we obviously want to go to a
bunch of other separate effects tests and see just how
well does it do with those as well, different
geometries.
This is the same calculation and the only
difference here is that little term I say we put in,
here's the influence of light gas accumulated in the
boundary layer. And the only difference between here
and here is that term, and this a hydrogen -- or
excuse me, a helium accumulation of 1.7 percent.
CHAIRMAN WALLIS: This is the average heat
transfer coefficient?
MR. HENRY: Yes, it is. That's all they
measured in that test.
Now I should -- I mean, to give the author
credit, he developed his own correlation for what that
was. This effective -- of course, this really should
just be heat transfer coefficient here. That's my
fault. But in essence, he had his own correlation.
Again, following in the structure of the code he
wanted to put something and clearly understand how the
code's using it. That's why we put in our own
correlation for it here, because we know exactly how
the information is getting transferred from node to
node to node.
But here you can see the obvious
influence. If you have a one node model, so we're
always sitting at some kind of mass fraction down here
someplace -- let's see, I should be more like in here.
Here. As opposed to pushing air out so some nodes may
be condensing here, but the break node is much more
down here. That has a tremendous influence on the
peak pressure that you would calculate.
CHAIRMAN WALLIS: So this noncondensible
mass fraction appears as N in this Fm?
MR. HENRY: NFST, that's all one thing.
CHAIRMAN WALLIS: Your also influence as
FST?
MR. HENRY: Well, NFST is the mole
fraction of steam. NF is mole fraction and ST is
steam.
CHAIRMAN WALLIS: So this is saying that
F is one plus something that's proportional to mole
fraction?
MR. HENRY: Yes.
CHAIRMAN WALLIS: And the Nussel number
goes up when N goes up or does it go down?
MR. HENRY: The Nussel number goes up with
increasing steam mole fraction. The more steam we
have in there, the more -- the measured -- yes, this
is N for mole fraction and F and ST is steam.
CHAIRMAN WALLIS: We're looking here at
noncondensible mass fraction.
MR. HENRY: Right. Since it's only air
and steam you could --
CHAIRMAN WALLIS: Use it the other way
around?
MR. HENRY: Right. But this is of steam
here. I could turn it around, but that's the way the
experimenter reported his data and I always try to be
faithful to what he represented as information.
MEMBER KRESS: Well, how did he
extrapolate against the cube of MC delta P of the
water, probably, you've got the area.
MR. HENRY: Measured wall temperature --
MEMBER KRESS: Measured wall temperature?
MR. HENRY: And the environment
temperature.
I just want to make sure, this was no
indictment of his correlation, but we put it in in our
own way and we know how the code's going to use the
information.
CHAIRMAN WALLIS: Well, this looks a
little strange, but I guess we've got to go on.
Usually when you put in a little bit of
air it has a big effect, and this looks as if it
doesn't. As a matter of fact, it's rather a gentle
effect of putting in air. You have to put in a lot of
mass fraction.
MR. HENRY: Well, we're going to get to
that, Graham.
CHAIRMAN WALLIS: Because there's a zero,
and you may never get to zero.
MR. HENRY: Right, never get to zero
there. Right.
CHAIRMAN WALLIS: Because zero is way off.
MR. HENRY: Right. We're going to get to
that.
MEMBER KRESS: I think that's the reason
we didn't get it the first time.
CHAIRMAN WALLIS: That's why I didn't
understand it because Fm seems to be linear and steam
fraction they always kind of leap up when you get very
close to 1.
MR. HENRY: Here's a couple -- Virgil
asked me a question before about what happens with
vertical -- with the flat systems, and in particular
the ones that are important to us are the downward
facing systems, which are the containment doom as well
as all the floors of the compartments. And that's why
we focus on Anderson's experiments because he had,
indeed, measured things, which I'll show you.
Let's go to his configuration, which were
interesting to us so that we could relate what he
measured in downward facing systems.
So he ha something that looked like the
top of the containment all the way down to the side.
So this was heat flux zone 1 up through 14. That goes
from vertically downward all the way up to -- excuse
me. Horizonal facing downward to vertical. And it
was a slice of the containment-like geometry.
MEMBER SCHROCK: Is the water running down
or is it dripping off?
MR. HENRY: Both.
MEMBER SCHROCK: Both.
MR. HENRY: Both. And the net result of
what he saw as we go from 1 to 14 as shown here for a
particular test, so this is horizonal facing downward.
This is cooling plate number one. But here's number
14, so this is the one that's vertical.
And what he saw from the practical point
of view is that there's no difference in the energy
transfer rate to the wall. And he had two different
ways of doing it with a heat flux measurement and a
containment energy balance here. And I need to get
back into his thesis to make sure I understand what
these -- the relative uncertainties of these are, but
that will come.
And what we gleaned from this is for
downward facing systems there's virtually no
influence. And, as you know, of course, when you go to
those natural convection kind of relationships, the
length essentially cancels out of it anyway.
So, this is just a preview of how we're
going to look at it, but at least this gave us -- and
this is things that Anderson reported -- of various
hot flow of temperatures and wall temperatures, this
is what Anderson measured as the heat transfer
coefficient. And this is what Dehbi's correlation,
which I mentioned here the author had formulated
himself, shows.
As the temperatures increase, which means
the pressure has to increase. These are reasonably
close. If anything, Anderson's tend to be higher than
Dehbi's or even more energy -- higher heat flux,
higher heat transfer coefficient than what we're
doing, except at this very low one. So this will also
dictate when we finally get to doing this detailed
comparison what the uncertainty boundaries are that we
think have to come from separate effects tests.
MEMBER SCHROCK: So do we know where these
experiments, where the relationship between the
resistance in the diffusion layer is compared to the
conduction resistance in the film?
MR. HENRY: For most of these that relate
to design basis type of energy transfer rates, there's
hardly any resistance in the film. Resistance is on
the gas side and/or on the concrete wall, and the film
is a very small amount of the resistance. We
struggled with that for a long time, Virgil, ourselves
and we went to the trouble of making sure that we had
this Laminar to turbulent film transition. We saw no
influence of it, but I'm not surprising you with it
I'm sure, you've seen it many times.
MEMBER SCHROCK: This funny shift from a
lower value in Anderson to a higher value in Anderson
as you go across these conditions which correspond to
higher temperatures in the steam environment.
MR. HENRY: Let me tell you where this
comes from. This comes out of Anderson's paper that
he put into literature. This table was in there --
MEMBER SCHROCK: I haven't seen it. Where
was that published?
MR. HENRY: I can get that to you. I'm
trying to think. I think it was in Nuclear
Engineering and Design. But I will get it for you.
MEMBER SCHROCK: And this thing you showed
us --
MR. HENRY: I've got to get his thesis so
I could understand where these numbers actually come
from, because he's obviously averaged over some of
these plates.
I'm sorry?
CHAIRMAN WALLIS: On the previous
transparency something went by me. You've got plates
at different orientations, is it related somehow to
the picture in slide 24. What were the various plates
here?
MR. HENRY: This is looking at a frontal
view of his experiment. This is the side view. So
these plates are individual plates that have their own
cooling core so they can --
CHAIRMAN WALLIS: This is like a sort of
two dimensional containment?
MR. HENRY: Yes. Steam comes in here and
they measure condensation rates in each one of these
plates under average conditions that are in that
table.
CHAIRMAN WALLIS: And the orientation
makes no difference?
MR. HENRY: The orientation makes very
little difference.
CHAIRMAN WALLIS: This might indicate that
it's some sort of a circulation locally that's been
happening rather than --
MR. HENRY: It is spinning.
CHAIRMAN WALLIS: Yes. Is that what's
happening or is it --
MR. HENRY: I really think that -- and,
again, I want to get his thesis so I understand more
than what's in that particular paper. But there's a
couple of things that have been going on here.
Obviously, you have heavy over light. But
if you collect enough water in this region, which is
just horizonal facing downward, that by itself is
going to fall away --
CHAIRMAN WALLIS: And it drips off the
top.
MR. HENRY: Drips off and that certainly
tears up any stable boundary layer. And what you
eventually get to over here, which is vertical, this
is also in excellent agreement with Dehbi's vertical
experiments.
CHAIRMAN WALLIS: This is with a lot of
noncondensibles.
MR. HENRY: Right, this is.
CHAIRMAN WALLIS: So it's nothing to do
with Nussel's film, and the limiting thing is in the
air spout.
MR. HENRY: Right. Right. And to that
effect, I should also mention they went to great
trouble to make sure the condensing plates weren't
limiting, for obvious reasons.
But I found this to be very helpful in
going to these containment conditions.
CHAIRMAN WALLIS: When this happens,
there's some global replacement variable, which is the
same for horizonal and vertical it's dominating
everything. Gravity doesn't really matter in that in
that global picture.
MR. HENRY: That's right. It's probably
a mixture of setting itself up this way as well as
stuff coming down this way, and so gravity doesn't
matter much.
But this also gives you the kind of
information you need to say, the containment side,
they're pretty complex geometry, how do I treat this
thing. And fundamentally what we say and our logic is
the length already canceled out anyway, so from a
practical point of view, systems which are facing
downward we treat with the same kind of heat transfer
coefficient, effective heat transfer coefficient,
because that's what these are all put in. HTC is heat
transfer coefficient.
There are a couple of things we should get
to, so I'm going to -- I do need to leave here not too
long after 5:00, Mr. Chairman, if that's okay.
For the Hitachi experiments and for the
Uchidas, I put -- this came, again, from the Hitachi
paper that shows their measurement. And they had a
geometry that was related strictly to suppression
pool. But again they were measuring the effective
heat transfer coefficient. Graham, the only reason I
put this up here is, here's your steepness that you
were looking for. So that's all there.
And this is not my line, this is
Hitachi's. However, the way in which MAAP looks at
Uchida, which is shown on that Hitachi slide is shown
here. So, that representation I showed you with Fm
etcetera as a function, and now this is the ratio of
noncondensible gas to steam. Here's the steepness and
this is the way that correlation looks. And this
needs to eventually have those same uncertainty
boundaries put on it, but this was all we could do is
digitize the information that came out in the original
Geneva paper, which was a real tiny figure.
I'm going to skip the next ones because I
want to get to some of the integral tests, because our
whole process is to try to build the understanding
from separate effects tests and then test their
capabilities when we get to integral experiments. So,
if it's okay with you, I'm going to jump to the CVTR
experiments.
So, this is CVTR, which is a
decommissioned containment now. I'll wait until
everybody gets this. And they had a line from an
adjacent power plant that came into here and it
discharged into this node.
Now, a couple of things here. This is a
12 node model and as we talked before, this is a
generalized containment scheme, so historically these
nodes got added later down here. That's why 11 and 12
are down in here. And 9 and 10 are embedded nodes that
are inside of the -- that we represent the refueling
cavity and I forget what else inside. It just doesn't
show up on this figure.
CHAIRMAN WALLIS: There are structures and
things in there that you don't show?
MR. HENRY: Right. And that's part of the
problem is, it's hard to find a description of all
those structures. But there is in the experimental
report, there is a specification of what the heat
sinks are and the uncertainties that they subscribe to
their estimation of heat sinks. So we use that. And
to some extent we have to do a little bit of guessing
of where they are, but there's only a couple above the
operating floor. This particular thing had a steam
generator on it. There's a fair size structure up
here.
MEMBER SCHROCK: When these guys do it for
the plant specific, they have to make these choices?
MR. HENRY: Right. They have to go look at
where they have rooms.
MEMBER SCHROCK: Yes.
MR. HENRY: And they certainly have to
have something which says I want to make sure that I
can see stratification if it would ever occur.
MEMBER SCHROCK: But you make a comment
that it's hard to come by that information.
MR. HENRY: For CVTRs it's hard to come by
that information, because it's a decommissioned
containment.
MEMBER SCHROCK: Okay.
MR. HENRY: In fact, it's being torn down
now.
The reason I wanted to make this point
here, there is a generalized nodalization scheme. You
could hook nodes together anyway you want, so node 4
can talk to node 11, there's no sequential problem
associated with it.
The other thing I wanted to show you was
the thermal couples that we will talk about here in
comparison, there's a thermal couple 28 that sits out
in this location, I think it's at elevation 370.
Thermal couple 11 is right below the operating deck,
so it's in this region between these two nodes.
Thermal couple 7 is sitting at, and I think it's
something like 297 or so, it's right here. And
thermal couple 5 is here. And just so you know where
they are inside the nodalization scheme.
And one of the things I need to show the
staff and of course this committee in the future is
suppose we started with one node, what would we get?
If we had two nodes, what would do we get? If we have
four, what do we get?
And also, I can tell you ahead of time,
basically if I would have made this one node, two
nodes, three nodes, four nodes, I'd get something very
close to what you see now. But don't take my word for
it. I owe that to you in the future.
We used these 12 nodes because we wanted
to see what are all the axial temperatures and the
influences on containment pressurization. So, we'll
keep coming back and forth to this, I'm sure.
There were three tests; test number 3,
number 4, number 5. The only difference is test 3 had
no sprays at all in it, so steam went into containment
and then it just cooled down over a number of hours.
In test number 4 they turned the sprays on
at about 210 seconds at half the capacity that the
containment had.
And test number 5, exactly the same except
full capacity of the sprays.
So this is the pressure that's measured
for all these different gauges throughout the volume.
And, of course, they're in very close agreement, which
is expected. And this is only the first 400 seconds,
this is the same set of measurements over the first
hour.
Remember that thermal couple 28 that's up
somewhere around 370 or so, that's this thermal couple
TC28 and showing here both the temperatures in node 1
and node 2, and this is that measurement for zero to
400 seconds and zero to 4000 seconds.
The sprays come on just about right here.
You see a little kink right there, and that's when it
comes on. And so all this that you see here is all
being driven by the spray cooling.
MEMBER KRESS: The break is when the steam
quit going in?
MR. HENRY: Yes, that's exactly right,
Tom, the steam -- the mass energy stopped right here.
Now, why we do this to begin with. We
thought MAAP4 was a pretty good code, as it
generalized nodalization and it had the kind of energy
with the natural convection thing I showed you that
didn't have Fm in it, and it allowed air to be pushed
around containment. And we did the best job we could
with MAAP4 and this CVTR test, we had a pressure that
was up here. It over-predicted the pressure by about
7 psi, as I recall.
The best nodalization we could think of,
all the heat sinks, everything else, the best thing
that we could put in there. So that's what really got
our attention. What are we missing? And the thing
that we're missing is when we do mass and energy
balances, as an example, we don't end up having any
idea of what that turbulence, whatever that
circulation is because we never were solving for it.
That momentum just disappeared. So everything that
was driving through the containment was all just due
to through flow and what it had to have to pressurize
the various other nodes to the same pressures. And it
would push air either way. But no way could we get it
down from up here to there. So that's where the whole
concept got started: What does this mean? What are
we missing in these nodes?
The other aspect is the temperature. We
over-predicted this temperature by something like 50
or 60 degree Fahrenheit. So we obviously had some
things that were really missing in both. What governs
the peak pressure, what determines the temperature in
these nodes.
And we also looked at the rest of the
temperatures as we worked down into the containment.
So this one is TC11 you see here, which is right below
the operating deck by 4 or 5 feet. TC7, which is
further down. And you can see with this one having a
peak of something in the range of 230, it's not too
much different than right above the operating deck.
When you get further down, this is hardly
increasing at all. And that's still a challenge to us
because this particular rise that you see right here
is only because the system's pressurizing because the
pressure is going up.
So what we do in order so that we have
some kind of perspective of what's going on, these
three lines right here are a heat sink that's a
quarter inch -- assumed to be quarter inch thermal
couple, which is a big thermal couple, just sitting in
that node. So how much would we slow down this
measurement if we had this generic thermal couple
sitting in there, because we don't know what that
thermal couple or RTD looks like, that wasn't in the
report.
Now you can see, that slows it down a
little bit, but still not as much as -- there's
something else even going on that makes that lower
region even cooler. But if we didn't have this
turbulence circulation velocity, we would really
overstate this temperature again, and this would also
be overstated because the pressure's higher. This
whole thing is coming about because the pressure is
going up and it's just eV to the gamma as a constant.
You can see certainly after the sprays
come on, we get quite bit agreement down low in the
containment as well.
And then we go to the very bottom of the
containment.
This is TC5 for the first 400 seconds, the
first 4000 seconds. And now we get much better, at
least understanding that this could be because of some
thermal response to the thermal couple and maybe it's
seeing some water dripping down from the wall.
This is again the average temperature up
in node 2, but the key thing I wanted to mention to
you the CVTR provides is it has detailed
representations of the liner temperature in the break
region.
So this is the side of the liner that
faces the gas space. This is the side of the liner
that faces the concrete. And in CVTR the liner does
not contact the concrete, at least not at the
beginning of the test. It's separated by 3/8th of an
inch. It doesn't mean that it couldn't be pushed out
during the test.
But this is our evaluation of the liner
temperature that's facing the steam. And the reason
this data that's shown here and the data shown here is
exactly the same, and the whole reason is that we
don't know where that measure was taken. We just know
where its elevation is. We don't know azimuthally
where it was in the test report. This is that
particular heat sink, which is our break node, which
was node number 2. And this is the node right beside
it at the same elevation. So we show a little bit
higher temperature, of course, in that node than we do
here, but at least we can see it's certainly following
the liner temperature quite well, which is one of the
evaluations that these guys have to do. They have to
evaluate the liner temperature during these design
basis calculation.
So that's why this was particularly
important to us. And this one I'm going to put in
better context for you in the future, but this is,
again, the liner temperature. We have nodes in the
concrete, which can be fairly thick. This is our
node, and this is what their temperature is, imbedded
in the concrete. And, again, these two are exactly
the same thing. It's just that they're two different
nodes at the same elevation.
I will put this in a heat flux context for
you, so you can really see this in terms of how much
energy, what's the transient deposition of energy into
the concrete, because that's really matters.
CHAIRMAN WALLIS: I don't quite understand
the lines here. The data are the results.
MR. HENRY: Right.
CHAIRMAN WALLIS: And there's something
called MAAP calculations -- which line is that? There
are two solid lines --
MR. HENRY: This is best estimate or what
I should be calling realistic just to get --
CHAIRMAN WALLIS: Realistic and
pessimistic.
MR. HENRY: Okay.
CHAIRMAN WALLIS: And then the data is way
down below there.
MR. HENRY: Right. This calculation right
here is basically the same as what's up here, because
this is the inside liner temperature. There's the
liner, then there's a gap and there's the concrete.
CHAIRMAN WALLIS: And your thermal couple
reading is way down there?
MR. HENRY: No, this thermal couple is
sitting in the concrete.
CHAIRMAN WALLIS: And what's the other --
MR. HENRY: This is our first concrete
node. The node can be -- so that's why I say in the
future I'll put this into the heat flux rate.
CHAIRMAN WALLIS: Somewhere in between.
MR. HENRY: Right. I'll characterize the
transient heat flux, which is more meaningful for you.
I apologize for that.
But this, of course, is easier to
represent. Okay. They have temperatures in the
concrete, how well you're doing there.
So these CVTR tests are very important to
us because that was the first clue we had there's
something that's really missing in this process and
what is it. Unfortunately right after that are the
spinning cylinders.
There's a couple more here that I'll go
through very quickly, again, Graham, with your
permission, just because this is a particular
containment configuration that gets put together two
different ways for these two CASP experiments.
CHAIRMAN WALLIS: It's very easy. We do
this for homework. I mean, you could just take one of
the rooms with flow in one and up the other side and
do some of that room calculation, you would show that
these flows in and out set up separation cells in the
room; they wouldn't be quite like your cell, but they
would be straight up.
MR. HENRY: That's right.
CHAIRMAN WALLIS: And you could actually
predict from some of CFD calculation what the role of
heat transfer should be. That would be not too
difficult a thing to do.
MR. HENRY: You got my attention. I'm
sure we'll come see you again.
CHAIRMAN WALLIS: Students do this for
homework.
MR. HENRY: I guess we'll have to find
someone who's younger and quicker.
CHAIRMAN WALLIS: Well, the frequency here
has CFD capability. It could do the same thing.
MR. HENRY: Sure.
CHAIRMAN WALLIS: That might be the more
realistic thing you should put in the cylinder.
MEMBER KRESS: Well, the building in some
ways you have to validate the concept he is trying to
put across.
MR. HENRY: And we'll go look for some
things.
I also apologize, I skipped over the first
HDR experiment very quickly, which is a large break
LOCA to get to CVTR, which is more meaningful. The
reason I skipped over HDR, not that it doesn't mean a
lot, it does. But as I showed you earlier on with 1
and 5 nodes, there was no benefit to looking at
circulation or turbulence or anything else. MAAP4 did
a good job with HDR. But the CVTR it stunk, so we
wanted to get right to the heart of the issue. And
the reason was we believe we were not correctly
representing the potential for energy transfer of the
break nodes.
I should also mention the way we do this
calculation of turbulence, etcetera, we get about the
right kind of circulation velocity that was observed
in CVTR, which is a very difficult thing to measure.
They did have some -- I think they had turbine driven,
turbine flow meters sitting in the annulus. You get
down to the bottom of the containment, their velocity
is almost none existent. So it's only a couple of
nodes that see this enhanced energy transfer rate.
Okay. The reason I want to touch briefly
on these, D15 with CFP1 with this schematic -- which
was again, now, Graham, this is their schematic, not
mine.
CHAIRMAN WALLIS: I realize that. The
actual thing looks quite different.
MR. HENRY: Right. This looks like it's
a straight through thing, which is what its intent
was, but when you get to the real thing -- I guess we
already went past it.
CHAIRMAN WALLIS: Did they clear the
special building for this test?
MR. HENRY: Well, this is a whole series
of tests. This went on for a number of years.
CHAIRMAN WALLIS: -- a series of
compartments.
MR. HENRY: So this says, for those of you
that may not have seen this before, it's breaking your
node out of room 6, and then it goes through room 4 to
room 8, then up to room 7 and into and out of 4.
These two roles are in line, and then into room 5 and
into room 9. So it looks all straightforward there.
But just so you appreciate the complexity of it, when
you look at the configuration, room 9 which is shown
here, includes all this annular region here, which is
there also, as well as this big hole in the middle
right here, which is this hole right there.
So here's room 6 and this is the break
pipe coming into it. And room 6 then flows through
room 4 here, the level of path, and this is room 4,
that little tiny thing, but it is the full height.
And goes into room 8. At room 8 goes up into room 7,
which is right above it. And room 7 over to room 5.
Here's 8 to 7. And back through 4 into room 5 and
then up to 9.
So it's a very complicated structure, but
it at least gives you an idea -- gives you a test of
how well you're doing representing the pressure
distribution in this particular test.
I also wanted to mention to you that there
are the two experiments; the test configuration of
course comes from the test report. The mass energy
releases and their uncertainties are characterized in
the individual test reports. So we used this.
The additional information is, and this
got us in touch with Teja Kantzleiter who was the key
experimenter on this a long time ago. He was kind
enough to send an email that defined the inner
surfaces of the outer concrete walls, the thing that
defined room 9, to have a 1 millimeter surface
coating, that the inner walls had half a millimeter
surface coating on both sides. So all those floors
and ceilings, and that had a thermal conductivity of
about .3 watts per meter degree K. And that the
concrete itself, of course, is density to specific
heat, and thermal conductivity to the best they could
figure out was about 2 watts per meter degree Kelvin.
So the information that you have in front
of you -- and again to try and get the most
information to you, these are fairly small figures --
but this represents the transient pressurization for
the most realistic behavior in containment. And where
this says optimistic and pessimistic, the pessimistic
also has in it their maximum mass and energy release.
The maximum you can get from that uncertain analysis.
And the optimistic has the minimum here, whereas this
which is realistic is using what they thought was
their best estimate of how fast this came in the
containment. And, of course, these are measured with
-- it's a two-phase flow and they're estimating it
from a momentum measured on the drag disk.
So this is the pressure in containment.
Temperatures for the three different things, again, in
these rooms close to the break. And this one is the
break room.
And this looks like a real mess here, like
a bunch of spaghetti, but what's shown, again, is this
generic thermal couple. So if we look at the solid
line, which is right here, as an example, that's the
most realistic representation, and this is that
generic thermal couple that we respond, and then right
above it here.
So, again, I don't know what their thermal
couples look like. I don't know if they're close to
any structure, etcetera, but at least we can see that
something is -- it's roughly a quarter of an inch
piece of structure holding the thermal couple in place
-- is one of the reasons these things could lag and
then the temperature could stay up. Because out in
here there is basically no motion going on. It's just
radiation, the environment, and natural circulation.
The blowdown's all over with back in here. Obviously
the blowdown is over with right there.
And this because we had -- we had talked
before about measuring the pressure differences
throughout the containment, these are now compartment-
to-compartment pressure differences. So now from room
4 to room 7 this is the measured pressure difference.
Again, this is in terms of Pascals, of course, but
it's negative because of just the direction of the
flow. But this is from room 7 to room 8.
CHAIRMAN WALLIS: Well, the Pascal is
pretty small.
MR. HENRY: Right, these are fairly small
pressure differences.
This is 10 to the 4th here that we're
looking at --
CHAIRMAN WALLIS: Oh, there is a 10 to the
4th.
MR. HENRY: Right. That's still not a big
pressure difference.
And the other point I wanted to make to
you, this one that was measured to be zero because
there are things we still want to make sure that the
code comprehends but does not comprehend -- as I said
it's a work in progress -- this is room 4 to room 5.
And that's that small little room where the holes are
in line. So in essence we get streaming flow directly
from 5 through 4 into 7.
CHAIRMAN WALLIS: And you don't deal with
streaming flow very well?
MR. HENRY: Right now the way the code
thinks of it, it goes from room 4 to 5 -- whichever
way this is -- and mixes and then it goes out. So it
needs a delta P to get out, that's why this is here.
But in essence it says there's no reason for me to
stop here.
MEMBER SCHROCK: Momentum never began and
got quieted down.
MR. HENRY: Right, and that is linear
momentum.
The only reason I wanted to show you that,
is there's another test, again, schematically now,
it's the same set of rooms but they're hooked together
differently. So now the break is into that little
room 4. Then it goes up and goes out those two holes,
which in 4 it was streaming through this way. Now it
goes out both ways into room 7, then into 8, into 5,
up into 9 and eventually comes around into room 6.
So this shows you, it's the same set of
figures now, but it's going into this little room here
4, so here's the pipe that's going into it, right
there. So it's this little square, but it is the full
height and this is a hatch on top which after the
experiment was over we detected there was a leak path
here from this break room into 9. So, again, we
include that in the representation. But as it goes
into that room, then it goes out sideways right here
into room 7 and 5, whichever way it was. One of them
went down into, I think, 8. Yes. And 5 went upwards
into 9 here and eventually came around and filled 6
from down below.
CHAIRMAN WALLIS: So if you have one of
these horseshoe shaped rooms, or whatever, I don't
know how you describe it.
MR. HENRY: Yes. Half a ring.
CHAIRMAN WALLIS: Half a ring around it,
you have the same circulation velocity in all parts of
it in your model?
MR. HENRY: Yes, there's the same
turbulence velocity in each node.
CHAIRMAN WALLIS: It's a first
approximation, right?
MR. HENRY: Right. And the key thing here
is --
CHAIRMAN WALLIS: I think a realistic
model would actually say we'll model the annular ring
as one thing and the cap as another. Two nodes
instead of one. Even though it's one room, but it's
so different.
MR. HENRY: Yes.
CHAIRMAN WALLIS: You're not going to do
that?
MR. HENRY: Yes, I can. I can.
And since you've made that point, I should
also tell you that clearly it represents this part of
this room 9, there's 2 nodes out here, 2 more nodes up
here --
CHAIRMAN WALLIS: There's all various
nodes that --
MR. HENRY: In the calculation, yes. But
I thought your point was in these also, because these
-- here's where the half thing is. You could
certainly do that.
One reason I thought this was also helpful
is the first was the linear progression through the
nodes. This is more like parallel flow paths.
And this is, again, the best estimate and
most realistic for pressurization in the break room 4,
and this is over the first 50 seconds, this is over
the first 1000 seconds. It didn't come through very
well, but that's 10 to the third here.
And then this is pressure difference from
room 4 to room 9 and the first 2.5 seconds. And the
pressure history of room 9 over the first 50 seconds,
which is one of the nodes in the outside region.
This is -- 7 to 8 there's an example,
which is break room into the next largest room out.
Excuse me. 7 is the next largest room out and then 8
is the room it goes down into.
CHAIRMAN WALLIS: When you've got this
concrete -- this insulating concrete wall -- doesn't
the thermal resistance of the insulation actually end
up dominating rather than the condensation side?
MR. HENRY: The only thing it's insulating
is the paint. The paint matters --
CHAIRMAN WALLIS: I thought they said they
had some coating on this.
MR. HENRY: Well, that's the coating, so
it's like an epoxy coating. And that epoxy coating is
only on the walls which are going outside, and it's
there to be a sealant.
CHAIRMAN WALLIS: But it is a significant
heat transfer, isn't it?
MR. HENRY: Yes, it is. And it is prior
to calculation.
We don't have to go through this detail to
compare. We just want to make that you can see that
it's doing a reasonable job on compartment-to-
compartment pressure history, transient pressurization
as well.
When we look at all these things, whatever
those various things that are happening in a point of
time, whether it's natural convection, forced
convection, etcetera, the uncertainty boundaries you
have for each of those models that came from separate
effects tests were the same in all cases. So you're
not tooting one of those parameters for a specific
test, and different tests.
The part which gets into uncertainties
gets to a short set of propriety slides, so I don't
know if we need to -- we can be out of here probably
about 15 minutes.
CHAIRMAN WALLIS: Well, you said this was
work in progress, so we're not -- you don't have to
give an evaluation of the MAAP in its final form.
MR. HENRY: No.
CHAIRMAN WALLIS: This is just to let us
know that you're doing it and get the feedback.
MR. HENRY: Get the feedback; I certainly
got plenty of that, and I appreciate it. And if
there's any experiments that you think that we should
have in this mix that we have overlooked --
CHAIRMAN WALLIS: When will this come up
in its final form?
MR. HENRY: We have a deadline to submit
to the staff in January, which is next year.
CHAIRMAN WALLIS: Fairly soon?
MR. HENRY: Fairly soon, yes. And then
the staff has heard from us twice on this; once in
June and yesterday to keep them updated on our
approach by the experiments. We want to make sure
that when it comes to the technical basis that we're
looking at things that you guys think, that they're
the driving force -- here's how I understand it -- it
must be doing, what the containment must be doing.
CHAIRMAN WALLIS: So you expect to come to
us again fairly soon with a finished product?
MEMBER KRESS: The staff review.
CHAIRMAN WALLIS: Or the staff has to
decide you want to do that.
MR. HENRY: Right.
CHAIRMAN WALLIS: They may not want you to
see us at all.
MR. HENRY: That's between you guys and
the staff. Certainly if you want us to come talk
about it, we're at your disposal to talk about it.
CHAIRMAN WALLIS: Well, it looks like a
considerable step forward in the modeling of
containment. Up there on the right-hand page
dovetails with industry.
MR. HENRY: Appreciate that.
MEMBER KRESS: Of course, the staff, they
plan, I guess, I don't know, access, too.
CHAIRMAN WALLIS: And I think also, since
this seems to be key for you, maybe that can be used
for these outbreaks.
MR. HENRY: Yes, we're particularly keen
on making sure that once we have a model that goes
with the experiments, that all that knowledge that's
associated with the experiments gets transferred into
their --
CHAIRMAN WALLIS: I think the outbreaks
are going to come to ACRS anyway.
MR. HENRY: Right.
MEMBER KRESS: Do you view this as saying
that in old code and this is a way to utilize that
margin by getting rid of some of those conservatisms?
MR. HENRY: That's exactly right, Tom.
But one of the ways that we would say that is that the
top suppliers do the right things for the right
reasons.
CHAIRMAN WALLIS: This margin isn't the
real margin, it's a margin of something in theoretical
equations, because you didn't know what was going on,
you had to have a -- when you know more you don't need
such a big margin. You probably mustn't get the
impression that they're somehow producing a safety
margin by producing an uncertainty which enables us to
make a better decision.
MEMBER KRESS: It's a some kind of level
of safety. I think you are reducing the margin,
because we're going to uprate the power and we're
going to put more stuff in, we are reducing the
margin. This just tells you you've got enough margin
there that you can do that.
CHAIRMAN WALLIS: I don't suppose you can
tell us what you mean by margin?
MEMBER KRESS: The difference between the
pressure and the design limit. The actual pressure
you get for design limit.
CHAIRMAN WALLIS: Actual pressure, not
just pressure.
So do you move on to the staff, then, or
do you want to say a little bit about this?
MR. HENRY: Instead of passing out the
proprietary slides, let me just say what's really in,
because that how we treat the uncertainties, what we
will eventually bring back to you. And what's
inherent in the process is that we believe that the
way you get closure is that you test against -- you
develop your uncertainties with separate effects
tests, and you work to these large scale tests for
closure. And by closure we're looking for the
realistic and the pessimistic and optimistic and we
try to stay away from conservative, because sometimes
we don't know what that is, given the attribute that
you're investigating. And we look to see if we can
bracket the data, not bound the data.
And once we're able to bracket the data,
we feel we have a 100 percent and 10 percent kind of
understanding of what's driving the bus and all these
analyses and also in the experiments. That's really
what's in the proprietary part of how we establish
that closure.
CHAIRMAN WALLIS: Is this congruent with
the CSAU?
MR. HENRY: Well, unfortunately I was part
of that once upon a time. It is consistent with that,
but -- and I was part of it when it was for direct
containment heating. And the only thing that's
different here from that is I tried to simplify it in
my own mind to fewer steps. But I also established
closure back to, say, you think you got this model,
are you able to bracket the data with the model, given
that uncertainties that comes from something else to
allow you to understand the detailed physics with the
processes that you're working or you just globally
bound it? And we would prefer to be able to bracket
it. These guys take it in-house, we want them to have
something that the engineers know where it all came
from.
CHAIRMAN WALLIS: You talk about the 95
percentile dosage, or just bracket?
MR. HENRY: We prefer to deal with just
bracket, but you certainly could take this to a
distribution. If you can do it with just bracketing,
well again the uncertainty bounds for individual
physics come from things like we saw with separate
effects tests. If you could live with that, you
shouldn't have to do anymore. If you want to look at
a distribution, you got to go back to those and define
the distribution and you put it into a Monte Carlo
kind of approach at a plant.
CHAIRMAN WALLIS: Do you have separate
effects tests of these circulation velocities?
MR. HENRY: You know I'm going to go look
for them. For the CFD calculations of flow into a
closed node, right?
MEMBER KRESS: This CVTR --
MR. HENRY: If I knew, I could find that
book called two phased flow, but I don't remember
those being in there.
MEMBER KRESS: Your CVTR containment
model, are there virtual boundaries in it as well as
--
MR. HENRY: Yes. I'm sorry, Tom. I meant
to make that point when we were there.
MEMBER KRESS: Yes. I thought that was
the one test where you really had --
MR. HENRY: I appreciate that. I'm going
to show you a couple. When we get to the HDR there is
a virtual boundary, but there's so many nodes, so many
rooms in the containment that you need to represent
all these -- or least virtually at least half of
these. And there's a boundary up here because at
E11.2 there was stratification.
When we get to CVTR, which I'm glad you
made that point, because those virtual boundaries are
here, here, here, here and this is treated as a
virtual boundary because I can't find out what the
grating was as you walk down. They're not very
specific about it and all the pictures are above the
operating deck. But this definitely is. These are
boundaries here and that is.
These are not -- obviously as you can see,
these aren't annular rings, these are just slices
through the containment. So, this is half of a
cylinder and this is half of a cylinder here.
MEMBER KRESS: How does MAAP deal with
creating stratification when you got light gas and a
heavy gas.
MR. HENRY: You can accumulate gas in the
node just because it eventually gets transferred up
and you slow down the condensation and slow down,
therefore, the energy transfer rate. Or you could
have it come in as it does in HDR at this kind of
location and it has a plume model that evaluates its
ability to mix if all this is really just relatively
quiescent system. Mix and rise to the top, but if
it's not completely mixed by the time it gets to the
top, it accumulates. And those virtual boundaries,
and even when we get to the plume model, that
entrainment rate goes back to the Recue Spalding
entrainment model, and then the kind of entrainment
coefficient that we use is defined by their model is
0.1, which is basically what they say to look at real
tiny gas-to-gas. But if you go look at volcanoes it's
roughly 0.1. It's the best estimate of the
entrainment rate of surrounding material.
Now we have bounding values on the other
side of it that are pessimistic and optimistic,
whichever the influence of the specific attribute that
you're looking at.
CHAIRMAN WALLIS: At 0.1 it's like you're
mixing when you get a plume that produces --
MR. HENRY: That's exactly what it is.
That's where it all came from. If we got something
that's an extremely powerful jet what's it doing.
MEMBER KRESS: In these containments, both
of them have sprays?
MR. HENRY: Yes.
MEMBER KRESS: If those are working all of
this gets overwhelmed by the sprays. The sprays do
everything. So this is only if the sprays are assumed
not to work?
MR. HENRY: No, the sprays don't always do
everything. But they eventually get into plant
specific analysis. But main steam line breaks, the
sprays do part of it but it's still pressurizing. The
only thing that turns around eventually is the M&E
stops.
MEMBER KRESS: What is the time for the
sprays?
MR. HENRY: The typical time for sprays is
anywhere from 45 seconds to a minute. But for main
steam line breaks, the M&E may last for 100, 200
seconds.
MEMBER KRESS: -- the time that you get
into the recirculating mode.
MR. HENRY: Well, we're still in the
injection mode, but it's still the sprays are not
necessarily turning the pressure around, they're just
slowing down its rate of pressurization. But the
spray momentum is also part of this whole thing here.
CHAIRMAN WALLIS: I understand some folks
have to go to the airport.
MR. HENRY: I appreciate that, Mr.
Chairman.
CHAIRMAN WALLIS: I don't want you to go
to the airport with too much momentum.
MR. HENRY: I'm going to spin out of here.
I apologize, but we do have to leave because I do want
to -- I will touch base with Rich, but the people from
the sites will be here also.
CHAIRMAN WALLIS: The people from the
sites are going to be here? I thought they were going
to leave first.
MR. HENRY: Excuse me. Tom Beach has to
leave.
Thank you for all your consideration.
CHAIRMAN WALLIS: That was very
interesting presentation and interaction.
MR. HENRY: I enjoyed it.
MR. LOBEL: My name is Richard Lobel, I'm
with the Plant Systems Branch in NRR I didn't come
prepared to make a presentation because the submittal
hasn't been made. There was question about how we
were going to proceed with the review, and we had a
short preliminary meeting this morning to talk about
that.
The review will be done in conjunction
with Research. In fact the Office of Research will do
most of the review because they, we felt, had the
expertise and the others and also had the resources to
do this. We wanted to make sure that we could do a
very thorough complete review of this, and the
expertise that's available in Research helps us do
this.
We will do contained calculations. We will
ask probably both licensees for the input to their
specific calculation in one form or another, whatever
is convenient for them and for us to use. When we do
an audit, that's usually how we work things out. We
have a conference call and ask them to submit it in
whatever form is convenient for the people here who
are going to be doing the calculations.
We also will be doing a little more of the
study of the uncertainties in the containment
experimental data. A lot of work has already been
done by Research, and that was another reason for
getting the Office of Research involved in what
normally would be just an NRR review. Because they
have a lot of expertise from work they've done in the
development of the contained code and comparing with
experimental data. And since Bob Henry didn't go into
it very much because a lot of that was the proprietary
part, but his method depends a lot on the use of
experimental data in the calculation of procedure
itself. And so we wanted to look in more detail at
the experimental uncertainty, too.
We haven't thought about it in a whole lot
more detail than that yet. We plan to do an
aggressive review when we get the submittal.
The plant specific submittals aren't due
until May. We're going to try to get the plant
specific information before the submittals are made if
that's possible so that we can start doing the
calculations earlier and identify the significant
issues as soon as we can.
That's about it.
MEMBER LEITCH: Have you used the MAAP5
before? In other words, are we looking at -- there's
two things we talked about was basically the change
from MAAP4 to MAAP5 and also the nodal concept. Has
the change from 4 to 5 been reviewed previously?
MR. LOBEL: No, I don't think we even have
MAAP5 in-house yet. They will be submitting that at
the same time. I understand from talking to Bob Henry
just before this session started that they will be
giving us a copy of that at the same time they make
the submittal.
MEMBER LEITCH: I see.
CHAIRMAN WALLIS: When you say a copy, do
you mean a copy of the -- the modern copy of the code
or you mean the documentation?
MR. LOBEL: No, the documentation.
Documentation.
CHAIRMAN WALLIS: Do you actually a
running copy of the code in electronic form?
MR. LOBEL: We may, and we may use that,
but we'll probably -- the plan is now to concentrate
more on using contained and comparing with their
analysis and let them run --
CHAIRMAN WALLIS: with other codes the
policy has been to endeavor to get an electronic copy
of the source code so you can run it.
MR. LOBEL: Well, we may do that and, you
know, we're certainly interested in your
recommendations and suggestions.
CHAIRMAN WALLIS: Well, we definitely
thought it was a good idea.
MEMBER KRESS: Well, this may be an
exception. MAAP I think belongs to EPRI and it's not
the licensee's code. It's not their privy to even
give it to the staff I don't think --
MR. LOBEL: But on the other hand, if we
really wanted that and considered that part of the
review, the licensees would have to try to accommodate
that as part of the review.
Let me say, a lot of this isn't going to
be a detailed review of MAAP. What we're going to try
to do more is review the method, because MAAP is a lot
more than just the containment. And what we were
going to try to do is -- the thinking is in NRR that
there's a couple of different options for the review
of MAAP that's still being talked about, as I
understand, in the office. And what we would do is
what we've been calling option one, which is look at
the models that are pertinent to the containment and
see that they're reasonable but concentrate mostly on
doing an independent analysis and a review of the
methods that are used in this procedure, which is a
lot more than just the code. It's their use of
uncertainty and experimental results. You saw that a
little of that from the pictures he was showing.
So, it's not going -- the plan was not
going to have this be much of a review of MAAP itself
except the specific containment models that are
involved and to concentrate mostly on audit
calculations and correlations, and that.
MEMBER KRESS: I presume this is a changed
licensing basis. Does that open the door for all
other PWRs to come in and do the same thing?
MR. LOBEL: It could, it depends on the
results of the review. What we've been asked to do
now is just to do the review of a general report and
then two plant specific analyses. But there was talk
at the June meeting about maybe having them come in
with a topic report that applied to more than just the
two plants. There wasn't any talk of that yesterday,
so I don't know what they're planning to do for that.
The broader the review is now, the easier
it will be on us in the future. We won't have to keep
going through this for four loop plants and ice
condensers and what else it may apply to.
CHAIRMAN WALLIS: Have you reviewed their
momentum equation formulation?
MR. LOBEL: No.
CHAIRMAN WALLIS: You've heard the
discussion here?
MR. LOBEL: Yes.
CHAIRMAN WALLIS: It would be unfortunate
if we had a code which seemed to work in comparison
with data but which had somewhat bizarre
interpretations of momentum balances.
MR. BOEHNERT: Extraordinary.
CHAIRMAN WALLIS: Yes. I'm sure Bob Henry
is smart enough to fix that up, but what appeared here
looked very strange. Maybe we're just being stupid.
It just looked very strange. We don't want to get
into a situation where something seems to work but the
theoretical basis justification doesn't really stand
up.
MR. LOBEL: Well, I think that we all can
agree that the phenomena is there certainly --
CHAIRMAN WALLIS: For other reasons than
the way that the math is actually sort of encoded in
the momentum equation. Maybe that the phenomena going
on which caused it are well represented by the way
things come together. And then maybe then it's up to
the person to bring together the documentation to give
a technically believable justification then for what
they do.
MR. LOBEL: The philosophy we've used in
other reviews is to try to not get in a position that
you were just talking about where the code may be
predicting data but have something in it that isn't
physically real.
CHAIRMAN WALLIS: That's the last thing
ACRS wants to have to fight regarding the --
MR. LOBEL: I guess we've already answered
this a little bit, I thought it would be worthwhile
for them to come and give you a presentation because
this is so new. It's a completely different approach
than what's in the standard review plan now for the
most part. We didn't have any plans to ask them to
come back again, but it sounds like to hear from them
after a point where we've gotten into the review
ourselves, so maybe a presentation on what they've
done and what we think of it after a round of
questions.
CHAIRMAN WALLIS: Then you've got the
submittal. They're going to have much more detail
about the technical basis because, again, what we've
seen so far doesn't really explain it well enough.
MR. LOBEL: Yes. I'll share that with you
if you want to see the submittal when it comes in.
CHAIRMAN WALLIS: Does the committee have
some other points at this time?
So what we need is just to -- the full
committee meeting we need an oral presentation --
MR. BOEHNERT: We make a subcommittee
report, or you're scheduled to make a report.
CHAIRMAN WALLIS: -- progress and that we
have some questions.
MR. BOEHNERT: Yes.
CHAIRMAN WALLIS: And I don't think we
need much time with the full committee.
MR. BOEHNERT: I think we've got a half
hour scheduled.
CHAIRMAN WALLIS: All right. We're going
to make it on time unless Professor Schrock has a lot
of questions.
MEMBER SCHROCK: No, I'm going to my taxi.
MEMBER KRESS: Now we know how to fix it
so Virgil doesn't have any comments.
CHAIRMAN WALLIS: Any reason why I should
not recess -- okay. I close the meeting, is that
okay.
MR. BOEHNERT: That's fine.
(Whereupon, at 5:27 p.m. the meeting was
adjourned.)
Page Last Reviewed/Updated Tuesday, August 16, 2016
Page Last Reviewed/Updated Tuesday, August 16, 2016