Thermal-Hydraulic Phenomena - July 18, 2001
Official Transcript of Proceedings
NUCLEAR REGULATORY COMMISSION
Title: Advisory Committee on Reactor Safeguards
Thermal Hydraulic Phenomena Subcommittee
Docket Number: (not applicable)
Location: Corvallis, Oregon
Date: Wednesday, July 18, 2001
Work Order No.: NRC-325 Pages 323-556
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
THERMAL HYDRAULIC PHENOMENA SUBCOMMITTEE MEETING
NRC-RES T/H RESEARCH PERTAINING TO
PTS RULE REEVALUATION
(ACRS)
+ + + + +
WEDNESDAY,
JULY 18, 2001
+ + + + +
CORVALLIS, OREGON
+ + + + +
The ACRS Thermal Hydraulic Phenomena
Subcommittee met at Oregon State University,
Richardson Hall, Room 313, Corvallis, Oregon, at 8:15
a.m., Dr. Graham B. Wallis, chairman presiding.
COMMITTEE MEMBERS PRESENT:
GRAHAM B. WALLIS, Chairman
THOMAS S. KRESS, Member
WILLIAM J. SHACK, Member
ACRS STAFF PRESENT:
PAUL A. BOEHNERT, ACRS Engineer
JACK ROSENTHAL, U.S. NRC, RES, SMSAB
DAVID BESSETTE, RES/SMSAB
STEPHEN BAJOREK, RES/SMSAB
VIRGIL SCHROCK, ACRS Consultant
ROY WOODS, NRC RES/DRAA/PRN3
NILESH C. CHOKSHI, NRC/RES/DET/MEB
JAMES T. HAN, RES/DSARE/SMSAB
A-G-E-N-D-A
Agenda Item Page
Loop Stagnation and Fluid Mixing in the
Reactor Vessel Downcomer Mechanics
for Primary Loop Stagnation. . . . . . . . 326
Cold Leg Thermal Stratification and
Plume Formation in APEX-CE . . . . . . . . 366
Downcomer Thermal Stratification in APEX-CE. . . 426
REMIX Calculations of APEX-CE Tests. . . . . . 436
3-D CFD Model of the APEX-CE Test Facility . . . 490
Summary and Reporting Schedule . . . . . . . . . 540
P-R-O-C-E-E-D-I-N-G-S
(8:15 a.m.)
CHAIRMAN WALLIS: This is the meeting of
the Thermal Hydraulic Subcommittee of the ACRS. We're
looking forward to hearing from Professor Jose Reyes
and his folks.
MR. REYES: Thank you.
LOOP STAGNATION AND FLUID MIXING IN THE REACTOR
VESSEL DOWNCOMER
MECHANICS FOR PRIMARY LOOP STAGNATION
MR. REYES: Yesterday we started by
discussing a little bit about the overall program and
we got into some of the integral test data.
Today we're going to focus more on the
separate effects or the local behavior in the
downcomer. So the presentations you'll see today all
deal with plume mixing and our predictions of plume
mixing or plume behavior in the downcomer.
And we'll also talk a little bit about the
-- this is kind of the bridge presentation here -- the
primary loop stagnation.
What we found was that some of the
behavior in the cold leg, as affected by HPI
injection, affects overall loop properties, in
particular, loop stagnation.
So we've identified the different
mechanisms for loop stagnation. So I will talk about
-- one of our tests is an inventory-reduction test.
And this is like being able to take snapshots of a
small break LOCA in progress. And it's similar to
some of the tests that performed the Semiscale. And
I actually compared some Semiscale data to our
facility just to give you a feel for where we lie in
comparison for that.
But that gives an idea of as we void the
plant and the steam generator tubes drain, at what
inventory would stagnation occur. And it coincides
with reflux condensation. So we'll look a little bit
more at tube voiding then.
Also in terms of one of our tests and what
we saw there, how tube draining would result in loop
stagnation for us.
And then steam generator reverse heat
transfer and then loop seal cooling. And I'll talk a
little bit about some of the downcomer behavior we
observed and how that affected the integral system.
The reason there's been so much focus on
stagnation in the past is that it was identified that
under stagnant loop conditions with HPSI injection,
that that would most likely be the most severe case
for producing cold plumes in a downcomer.
And I think today what you'll see is that
we've come to a slightly different opinion for this
particular plant, and we'll show you why.
So first the stepped inventory reduction
test. This test was performed basically by holding
the power constant. We had a constant steam generator
pressure, and we opened up a small break on the
reactor pressure vessel and we would drain some of the
inventory. We would stop the drain, and then we would
let the system go through natural circulation. And
we've measured the flow rates in each of the cold
legs. And we --
CHAIRMAN WALLIS: Do you have a break in
the vessel?
MR. REYES: Remarkable. It's just a small
little drain valve.
CHAIRMAN WALLIS: It doesn't sound like
one of those standard accidents, does it?
MR. REYES: No. It's not a -- this is not
a standard accident. This is a drain valve that we
use for setting up the plant.
So we would continue in step-wise fashion,
drain the plant. We stopped the drain. We let it
reach steady-state natural circulation conditions. We
go single phase and then we transition to two-phase
natural circulation. And then eventually we pass a
maximum and then to zero conditions.
And the same tests were performed in
Semiscale back in the 19- -- late '70s, early '80s
maybe -- a while ago.
MR. SCHROCK: So it's a series of
steady-state tests that reduced inventory; is that
basically it?
MR. REYES: That's correct.
So we start -- in this corner here, we're
starting with a system completely filled or including
a pressurizer, and we let it go to natural circulation
conditions. And so these are the flow rates, the
cold-leg flow rates that you'd see.
This is the sum of all the cold-leg flow
rates is the core flow rate. And you see for this
region here it's fairly flat. And what we're seeing
is just single-phase natural circulation.
We transition then. As we drain our
inventory on the bottom, on our x axis here, this is
a percent of overall reactor coolant, the primary-loop
inventory.
As we drain, we reach a knee here, an
inflection point, and now we're starting to go
two-phase. And we see an increase in our core flow
rates. And this is measured by the flow in the cold
legs.
And so what we found in this test is that
loop seals play a nice role of separating -- of
keeping the cold legs single-phase.
As we continued reducing our inventory, we
reach a maximum. And presumably the maximum coincides
with the point where you've essentially got two-phase
going up the tubes and essentially all-condensed
liquid coming down, or something very close to that.
So that would give you your maximum flow through the
core.
As we continue draining, then we start
seeing a decrease in the flow. These dark triangles
are steady-state points and the diamonds are actually
the transient data.
And we come down further and further as we
reduce -- somewhere around 60 percent of our overall
primary mass we reach a point where essentially we're
in reflux condensation.
And the measured flows then in the cold
leg are essentially zero. So injection during this
point would produce conditions similar to what's been
tested in the past: Cold injection with stagnant
conditions in the primary loop.
So our system was about 60 percent of the
overall inventory. We want to compare that to what
was done in the past with Semiscale. You can see that
the trends are very, very similar. We're just a
little bit offset to them. This has been normalized
to the same maximum there.
We -- somewhere they both around 65 to 70
percent -- my battery's charged -- somewhere around 60
to 65 percent of the mass -- of the total inventory --
now this excludes the pressurizer liquid mass. That's
why the number's different here.
Semiscale performed the test with the --
starting with their pressurizer essentially empty. So
they -- all their numbers were based on the percentage
of primary mass without the pressurizer inventory. So
that's what's been done here. So we have very similar
results for this design.
Okay. Well, that gives us an idea that if
our inventory during a small break LOCA drops to
about, in this case, 65 to 70 percent of the plant
inventory, of our initial inventory, we would expect
it to be in stagnant conditions.
MR. SCHROCK: There's an implicit
assumption that the inventory distribution is the same
in a steady-state reduced inventory situation as in a
transient. What transient, I guess, is the question
that comes to mind.
MR. REYES: That's right.
MR. SCHROCK: You have on the previous
graph a lot of data points designated as transient
data. What is the transient, just --
MR. REYES: Good. The --
Would you turn to the previous slide for
me, please? One more.
MR. SCHROCK: The one before that.
MR. REYES: Yeah, that's it. Thanks.
So what we're designating here is -- what
we were doing with these steady-state points, we were
holding at one position and just repeating the
measurement and coming up with an average condition.
And that's how the tests were performed at Semiscale.
When we looked at our actual measurements
throughout the test, as we were -- as we were
draining, we saw that our drain was slow enough to
where we actually could use the -- and I'm calling
this the transient data, because it includes the time
periods when we were actually draining the facility,
so the triangles represent the step changes.
We found that the transient data in
between the step changes, because the drain was slow
enough, seemed to match the steady-state data rather
well.
CHAIRMAN WALLIS: So this isn't some kind
of a standard transient. This is a transient that you
-- you -- you did --
MR. REYES: That's --
CHAIRMAN WALLIS: -- and you did it slow
enough so that it was quasi-steady state?
MR. REYES: That's right. So this is --
that's exactly right. So this is --
MR. SCHROCK: A small break line on the
vessel.
MR. REYES: Correct. So we're draining
from a low region in the plant.
And the idea was just to see what kind of
inventory behavior. So you're absolutely right.
Depending on the break location, we would see
different -- different behavior, different voiding of
the steam generator tubes. The void distribution is
important.
This is just a snapshot of a very
particular controlled test or controlled small break
LOCAs.
MR. SCHROCK: So your transient in this
case is -- appears to be always quasi-steady?
MR. REYES: That's right. That's right.
And originally we were just going to use
the steady-state data itself to compare with Semiscale
because that's how they performed their test. But as
we looked at the overall data we realized that --
actually the test was performed slow enough to where
we could use the transient data.
Okay. So that gave us -- at least that
gave us some idea of how the loop seals were
performing, and they were preventing bubbles from
getting into the cold leg, and at what conditions we
might expect to see some stagnation occurring.
So loop-stagnation phenomena. When we
talk about loop-stagnation phenomena what we're saying
is that the flow in the cold -- the flow through the
loop seals and up through the cold leg is essentially
zero. The flow rates are essentially zero.
You can have HPSI injection, which
produces some flow. And that will produce a cold
layer on the bottom of the pipe. And that concern, of
course, is that the cold layer spills into the
downcomer and it produces some plumes. And so you
might see it like this.
CHAIRMAN WALLIS: Now again in some of the
reactors the HPSI velocity is so great that the
momentum will carry it to the right.
MR. REYES: Correct. Yeah. So for
example for the side-injection B&W plant they have a
very high injection flow rate. It's a small injection
nozzle. And you'll actually jet across the pipe and
impinge the other side.
CHAIRMAN WALLIS: And it swells up on the
wall.
MR. REYES: That's right. So you get a
tremendous mixing.
This plant, for the Palisades plant, what
we found was that it was the opposite. They have a
very large pipe at fairly low injection flow rates.
So on a scale -- for their plant the maximum flow rate
was about 300 gallons per minute per cold leg of HPI
flow. So total injection flow was about 1200 gallons
per minute at the maximum.
CHAIRMAN WALLIS: So it makes a big
difference whether the stagnation is really zero flow,
or a little bit one way, or a little bit the other
way. I think that would make quite a difference since
your HPI flow is so low.
MR. REYES: That's right. Yeah, our HPI
flow is quite low. Yeah.
And there's a few things that we've
noticed that, again, we start to couple the separate
effects with the integral. And I'll show you what
happens with that.
CHAIRMAN WALLIS: I presume with this
Froude number criterion about whether any of it flows
back, and things like that.
MR. REYES: Right. Right. So the
criteria --
CHAIRMAN WALLIS: There must be some
condition where all the HPI goes into the vessel.
There must be enough. When there's enough flow rate
in the cold leg, then all the HPI flow will go to the
vessel itself. You've shown it sort of half and half,
or something like that.
MR. REYES: Right. This shows spilling
over this weird -- this little reactor coolant pipe
lip. And for certain conditions below a certain flow
rate all the flow will go towards the downcomer.
CHAIRMAN WALLIS: Does it then fill up the
loop seal and --
MR. REYES: About -- for higher flow
rates, you spill over and then you --
CHAIRMAN WALLIS: But then it eventually
fills up the loop seal with cold fluid, and then it
all comes back again, doesn't it?
MR. REYES: That's where -- that's where
I was surprised.
CHAIRMAN WALLIS: Hmm?
MR. REYES: This is where the data
produced a surprise for us, so I'll show you.
CHAIRMAN WALLIS: Okay. So we haven't
seen the -- this is just --
MR. REYES: Yeah, you haven't --
CHAIRMAN WALLIS: -- act 1, scene 1 so
far.
MR. REYES: That's right. The plot
thickens.
MR. SCHROCK: It looks like the plume
narrowed, though.
MR. REYES: The plume narrowed, but the
plot thickened.
These are the integral system tests that
we performed. And we wanted to understand a little
bit about the stagnation. One stagnation would occur
for these different types of integral system tests.
And so we had the -- numbers 7, 8, and 9,
these were essentially small break LOCAs -- well, they
are small break LOCAs. For 10 -- 11 and 12 are main
steam line breaks. And then number 10, which sits
right in the middle there, that was a combination,
stuck-open ADV, which would be like a main steam line
break, in combination with the stuck-open pressurizer
safety release valve. So it's a primary side break.
So it's a combination case. We wanted to see what
would happen with regard to stagnation.
What we observed in those tests is that
for the very small break LOCA, cold legs 1, 2, and 3
stagnated. And it stagnated because of the presence
of this cold loop plug in the loop seals, so it's kind
of interesting.
For Test Number 8 we saw a combination of
things happening as a larger break/small break. In
the previous test there was no tube voiding. In this
test there was tube draining.
So in the small break, the very small
break, what we found was that the HPI could keep up
with the break flow. As a result we saw the tubes in
the steam generator remaining full.
In the two-inch break we saw draining, so
we saw a combination of effects there. Cold leg 1 and
2 stagnated because of steam generator tube voiding.
And cold leg 3 and 4 stagnated a little
bit before each of the respective partners because of
this cold liquid plug forming in the loop seals.
For the main steam line breaks down below,
11 and 12, we saw 2 and 4 stagnating due to a loss of
steam generator heat sink, reverse heat transfer from
the steam generator. But we also saw some interesting
behavior here with regard to the downcomer.
The same thing with number 12 where we had
the break on the other side of the plant. Cold legs
1 and 3 stagnated due to the loss of the heat sink.
The combination, stuck-open ADV and
pressurizer safety relief valve, we saw 2 and 4
stagnate due to the -- because of reverse heat
transfer in steam generator 2.
So we observed different mechanisms for
stagnation for these different tests, which was very
nice because that allows us to characterize the
downcomer under different situations and plume
formation.
MR. ROSENTHAL: I'm sorry.
MR. REYES: Sure.
MR. ROSENTHAL: If you would just go back
one slide.
MR. REYES: Yes.
MR. ROSENTHAL: In all these cases the
steam generators are full?
MR. REYES: No.
MR. ROSENTHAL: Except for 10?
MR. REYES: In all -- so Test Number 8,
which was a two-inch break, in that case we did see --
MR. ROSENTHAL: I'm sorry.
MR. REYES: -- the steam generator tubes
drain. In the other test --
MR. ROSENTHAL: I meant the secondary side
of the steam generators.
MR. REYES: I'm sorry? The primary --
MR. ROSENTHAL: The secondary side.
MR. REYES: Oh, for the main steam line
breaks the secondary sides, of course, drain for the
-- for the broken steam generator. We let them drain
all the way out.
Now we --
MR. ROSENTHAL: Well, I mean the feed
water would keep up with the stuck-up in ADV, so these
are just postulated scenarios.
MR. REYES: Correct. Correct. That's
right. That's a very important point.
So some of the assumptions for the main
steam line breaks we assumed that the operator
continued to feed the broken steam generator for 10
minutes. And then we would adjust -- at which point
they would diagnose the situation and close the --
close the feed water to the broken steam generator.
And then they would work with the intact steam
generator to restore the heat transfer to the system.
For most of the cases, because we had such
a -- with such a -- in the main steam line break, for
most of those cases because we had such a rapid
cooldown, the feed-water flow and the steam flow on
the intact steam generator was essentially isolated,
because we just produced a very cold temperature in
the core.
CHAIRMAN WALLIS: Yeah. One steam
generator is quite enough to cool it down.
MR. REYES: That's -- that's right.
CHAIRMAN WALLIS: It's no problem at all.
MR. REYES: Absolutely.
Okay. So I wanted to just illustrate a
little bit of what we saw. I picked the two-inch
break because that had the draining of the steam
generator tubes, so I wanted to show that one.
It also had the -- the case of the loop
seal cooling, which resulted in stagnation. So I
picked this one. It's a little bit more complicated
to demonstrate, but I think it's reasonably clear,
but...
CHAIRMAN WALLIS: What you mean -- you
don't mean heat transfer from the loops. You mean the
cold fluid going into the loops?
MR. REYES: That's right. That's right up
where I mean.
CHAIRMAN WALLIS: Because I was really
puzzled when you had this had a loop seal coolant. It
means that it's got cold water in it.
MR. REYES: Right. It's -- right. It's
cold water mixing in the loop seal. Thanks.
So this is the behavior inside the steam
generator tubes for the two-inch break case. And what
we see is that this is -- we have a set of level
measurements, they're just DP cells in -- attached to
the long tubes, another set attached to the shorter
tubes. So we have the maximum and the minimum
basically for those and, as far as we can observe, the
draining in both of those, in both cases.
What we do see, and this is an area of
interest to us with regards to RELAP, is that the
tubes, first of all, they drain at different rates, at
different times. And that's what you'd expect.
What it does -- what it does mean, though,
is that if you're modeling in RELAP with a single
tube, the question comes up, well, which tube are you
modeling. Is it some average tube, or how does that
work.
So that's kind of interest to us because
as long as these short tubes are filled, you still see
flow in your primary loop. So there's a range here
from about 1500 seconds all the way out to, oh, maybe
about 3,000 seconds that you still see some flow --
CHAIRMAN WALLIS: Excuse me. Long tubes,
short tubes? There's some long tubes out here and
some short tubes in here, or something.
MR. REYES: That's correct. Yes. Long
tubes go all the way to the top of the steam generator
tube bundle and short tubes are the inner circle, the
inner tubes.
CHAIRMAN WALLIS: And you're scaling the
way a typical steam generator is in terms of the ratio
of the lengths, are you, or is it exaggerated in your
facility?
MR. REYES: So our facility, it's about --
the length ratio is constant. It's about 1 to 3.5.
So all of our lengths about 1 to 3.5 in the steam
generator.
So we're actually a little long compared
to -- compared to Palisades -- compared to our normal
one-fourth scale.
So that's what we see there. We see this
draining really begins somewhere around 1500 seconds
or a little bit earlier on the long tubes. But we
still see flow in the primary loop. And then we
continue to see flow until the shorter tubes start to
drain.
MR. BESSETTE: You see the same sort of
thing in ROSA at full height. Roughly, basically the
same.
CHAIRMAN WALLIS: It means that in RELAP
they might be useful to bundle the tubes.
MR. BESSETTE: In RELAP typically the
generators model a single tube.
CHAIRMAN WALLIS: Yeah, but that's --
MR. BESSETTE: We have done sensi- --
CHAIRMAN WALLIS: We miss that effect
altogether.
MR. REYES: Yeah. We have done
sensitivity studies where we model three tubes, or
whatever.
MR. BAJOREK: It depends on the sequence.
MR. REYES: So here's a look at the
cold-leg flow rates. And looking at cold leg number
2 and cold leg number 4 as a function of time for this
test.
And here I've identified when steam
generator tubes and the long tubes begin to drain, at
the very onset of draining the tubes. And then -- so
these are the two different flow rates. The one on
the bottom here, which comes along and -- they match
pretty well for this initial portion of the transient.
And then we see they kind of split off here. The cold
leg number 4, that goes to zero.
Cold leg number 2 experiences a small
increase and then continues to decrease down.
What's happening --
CHAIRMAN WALLIS: Where exactly are you
measuring this flow rate?
MR. REYES: These are being measured in
the cold legs.
CHAIRMAN WALLIS: Yeah, but where?
MR. REYES: Oh, so these are just after
the injection location.
Is that right, John?
MR. GROOME: Yeah. They're right after
the high-pressure injection --
CHAIRMAN WALLIS: You mean they're
upstream?
MR. GROOME: -- before the reactor vessel.
CHAIRMAN WALLIS: Oh, they're on the side
of the vessel.
MR. REYES: Right.
CHAIRMAN WALLIS: So --
MR. GROOME: They're on the downstream
side of the loop seal.
CHAIRMAN WALLIS: Yeah, but which side of
the injection point are they?
MR. REYES: So they sit --
MR. BESSETTE: The vessel side.
CHAIRMAN WALLIS: They're between the
injection and the vessel.
MR. REYES: Right.
CHAIRMAN WALLIS: So you could have zero
flow in the cold leg, and then you would just pick up
the injected flow --
MR. REYES: We would pick up some injected
flow.
CHAIRMAN WALLIS: -- which would be about
one gpm somewhere.
MR. REYES: Right. And then we have a
separate flow meter measuring our injection flow rate
also. So we know at all times how to --
CHAIRMAN WALLIS: When you say this goes
to zero, if you actually had injection --
MR. REYES: Injection flow.
CHAIRMAN WALLIS: -- it could be zero in
the cold -- in loop seal and then it would be the
injection flow where you're measuring it.
MR. REYES: That's right. That's right.
Yeah, we haven't subtracted those out.
CHAIRMAN WALLIS: Is that why it bottoms
out at one instead of zero?
MR. REYES: At one, yeah. That's right.
That's right.
So what we saw happening here, we were
interested in how this flow was splitting. We
expected that as the tubes drained, the flow in the
primary loop would just continue to drop.
Because we're draining 133 tubes, it just
kind of goes in a fairly well-behaved manner towards
zero. But we saw the split here, and we were curious
about how -- what was happening there.
And so what we observed was that this Weir
wall, there was spillover over the weir wall, this
reactor coolant pump lip. And the cold water would
create a plume basically and fall into the loop seal.
And now you're producing a very cold loop
seal. So the next figure here shows that, illustrates
that. So here both loop seals are warm, both for
number 4 and number 2.
When we get to this -- that point, that
transition point there, the 1500 seconds or so, and we
see that -- for the cold leg 4 we see some of the
plume spilling over into the loop seal and starting to
cool that off. And that corresponds exactly with the
time that we see that split.
So the loop seal forms -- it's basically
a cold-liquid plug which resists -- that gravity head
there resists the fairly low flow of the loop. And
the flow is diverted then to the other cold leg.
Through the common lower plenum of the
steam generator, the flow is diverted to cold leg
number 2. And so we see an increase in the flow rate
there and a zero flow in cold leg 4.
CHAIRMAN WALLIS: Did you get any kind of
loop seal blowout with these cold plugs being blown
out at some stage?
MR. REYES: We'll have to look at some of
the other tests to see.
CHAIRMAN WALLIS: That might not be too
good for the reactor vessel, blew out a plug of cold
water.
MR. REYES: If it put out some cold water.
MR. BESSETTE: You need larger breaks to
get a loop seal clearing basically in this.
CHAIRMAN WALLIS: You need a larger break
for that to happen?
MR. BESSETTE: Yeah, to create loop seals.
MR. ROSENTHAL: I'm not sure that the
vessel, given its mass, would even notice a slug.
MR. BESSETTE: Plus you've got to put the
break into the cold.
CHAIRMAN WALLIS: Well, it's a question of
local cooling, isn't it, and --
MR. BESSETTE: You've also got to put the
break in the cold leg. This was a hot leg break.
MR. REYES: Yeah, the thought was if --
CHAIRMAN WALLIS: It was just a thought,
I mean. While we're thinking about it, it's sort of
like the boron-dilution problem where you've got a
slug of something that comes in.
MR. SCHROCK: In the previous diagram you
have certain events highlighted, number 2, long tubes
begin to drain.
MR. REYES: Yeah.
MR. SCHROCK: Is that something that you
see from these traces, or that's just pointing out
that this is observed from other evidence at those
times?
MR. REYES: This was what was observed
from other evidence.
CHAIRMAN WALLIS: What's the big hiccup at
about 1100 seconds?
MR. REYES: We know that the initial steam
generator tube draining does result in an increase in
flow, so we're draining those tubes and they're
draining into the cold-leg side and to the hot-leg
side. And we expect to see some increase in flow
rate. I would suspect that's related to that.
CHAIRMAN WALLIS: Yeah. But how do you --
MR. REYES: And we don't see it in the
other one, so --
CHAIRMAN WALLIS: You need to blow-up the
scale to see for how long it stayed at zero at around
1100 seconds, if it did. You know you can't see it in
here. Eleven hundred seconds is a big hiccup.
MR. REYES: Oh, this -- this spike right
here you see. Yeah.
CHAIRMAN WALLIS: Yeah, it goes down to
zero, is the point.
MR. REYES: Right. Now the -- these
magnetic flow meters, they're very sensitive to
voiding. If you have a bubble through the line and it
goes past the instrument terminals, you might see a
spike like that.
However, I tend to think this might be a
real -- we'll look at that a little bit closer. This
might be a real...
So what we see is the creation of this
cold-liquid plug in the loop seals and the flow being
preferentially diverted to the -- through the steam
generator lower plenum to the other flowing leg with
the warm loop seal.
Later today we're going to show you how
this works. We'll take -- when we go to the lab we're
going to run this experiment for you. Here we have
the -- this Weir wall.
Imagine in the plant that this whole elbow
here would be the pump and this would be the outlet of
the pump. And you can see that it acts effectively as
a dam which keeps the cold water on the reactor vessel
side.
Eventually you will spill over, and you
will go into this loop seal, and you'll cool it off.
We'll show you that phenomena.
CHAIRMAN WALLIS: Now RELAP has great
trouble trying to predict that, doesn't it?
MR. REYES: Excuse me?
CHAIRMAN WALLIS: Doesn't RELAP have great
trouble trying to predict that, that sort of
phenomenon?
MR. REYES: I would say that there's no
way.
CHAIRMAN WALLIS: But you can still run
RELAP.
MR. REYES: That's right. And I think
what we saw in this test with the RELAP calculation is
that it did predict reasonably well the steam
generator tube voiding on some average basis.
And so it was in the ballpark. But as far
as the sequence of -- the timing and the exact
sequence, probably would miss. It can't --
CHAIRMAN WALLIS: It just mixes --
MR. REYES: -- it can't calculate this.
CHAIRMAN WALLIS: It just mixes
everything, doesn't it?
MR. REYES: Yeah.
We're looking at that, though, to see what
RELAP is predicting as a loop seal temperature during
the trading and comparing it to what we saw.
So you'll see that test a little bit
later. Chris Linrud and Ian Davis are here to do that
for us.
Okay. So we saw steam generator tube
voiding as a stagnation mechanism. We saw this cold
loop seal plug as another stagnation mechanism.
The third one that we observed was this
reverse heat transfer. And that's when the steam
generator and the main steam line break, we blow that
steam generator down on one side. The other intact
steam generator is bottled up and becomes a heat
source eventually.
And so you see in this schematic here,
this is for Test Number 12, a main steam line break at
full power. We have steam generator temperature and
the hot-leg temperature. We see initially the hot-leg
temperature is greater than the steam generator
temperature.
As we go through the blowdown, of course,
we drop that hot-leg temperature well below the steam
generator temperature. And we see stagnation of cold
legs 1 and 3 occurring within this band right here.
So it doesn't occur immediately. There's a band that
-- there's a delay time and then a band. And we were
curious about that.
Now later on what we see is the -- as the
primary plant reheats and repressurizes, our
temperature goes up above the steam generator
temperature. And we see a resumption in the cold leg
1 and 3 flow. So it works as expected. And again we
see that occurs relatively close to when we cross that
-- when we made that transition.
We were -- and that's to be expected.
What we were curious about was this little delay here
and what might be causing that. Here's looking at
the, again, cold-leg flow rates for 1 and 3. And you
can see that we start off with our pumps on, come way
down. We go to zero, essentially zero flow for the
flow meter.
Come -- and we resume our -- this steam
generator is a heat source. And here the steam
generator is a heat sink again. And we come back to
our flow intake.
CHAIRMAN WALLIS: What does negative flow
in the cold leg flow area mean?
MR. REYES: Negative flow, I don't think
we're seeing negative flow in this. It's just the --
CHAIRMAN WALLIS: Isn't that what it says?
MR. REYES: It certainly does, but I think
we're just zeroed. We're a little bit off in our
zero.
This flow meter is -- I'll ask John Groome
to tell me the range for that flow meter on the test.
Do you remember this one?
We've been struggling with memory lately,
myself.
MR. GROOME: Yeah, I'm sure my memory's
not any longer than yours though, Jose, so I don't
know if I'll be able to answer the question, but my
name's John Groome.
And these flow meters are ranged positive
to negative 100 gallons per minute. And, you know,
there's some interpretation that you have to do with
all data.
And I'm of the same opinion as Jose,
that's just a zero shift there on that second meter.
And that could be not just due to the meter. It could
be due to the actual estimate loop that carries the
signal back to the data acquisition, so...
CHAIRMAN WALLIS: You really need a more
sensitive flow meter because you're talking about low
flows in the order of one gallon per minute.
MR. GROOME: Yeah. And, you know, these
flow meters are actually quite phenomenal. When we
first worked with Westinghouse back in the early '90s
we couldn't measure flow in our cold legs. And so
part of the problem was our temperature and pressure.
And so we actually worked with them to
develop a new flow meter to -- that we put a flow
meter in to test, at about $4,000 a test we'd break a
flow meter. And so these flow meters have been in for
about going on three and a half years, and they've
worked for three and a half years successfully.
And, you know, this is a meter that's
capable of flow ranges up to about 700 to 800 gallons
a minute. So when you're down here looking at zero
and you're nitpicking about a gallon-per-minute flow,
I think you ought to be happy you have any data.
CHAIRMAN WALLIS: Well, how does a flow
meter work when you have, say, a stratified
countercurrent flow? Does it --
MR. GROOME: Well, it's a void average.
Unfortunately, it measures in gallons per minute. But
it's a void average flow, a volume-average flow.
CHAIRMAN WALLIS: So it has -- does a --
is it slices, or something, or --
MR. GROOME: No. It actually has just
two, and I can show you some flow meters today if
you'd like, but it just has two contact points in the
middle. So it assumes that the velocity profile
through the -- through this bore of the flow meter is
constant.
CHAIRMAN WALLIS: Well, that's very
difficult if you've got a countercurrent flow, with
cold water going one way and hot the other, --
MR. GROOME: Sure.
CHAIRMAN WALLIS: -- it must get very
confused.
MR. GROOME: Sure. And that could be
what's happening.
CHAIRMAN WALLIS: So it might well give
you a faulty reading.
MR. GROOME: Right. Right.
MR. REYES: Sure. Sure. Yeah.
Thanks, John.
CHAIRMAN WALLIS: So it just measures at
one point essentially?
MR. GROOME: Correct.
CHAIRMAN WALLIS: And then assumes a
velocity profile?
MR. GROOME: Correct.
CHAIRMAN WALLIS: Oh.
MR. REYES: So reduce the stagnation
occurring and then resumption.
CHAIRMAN WALLIS: So what it's measuring
is the velocity at the point where it measures what it
calls flow. That's what it's really doing.
John, it's --
MR. GROOME: Excuse me?
CHAIRMAN WALLIS: It's really just
measuring the velocity at the point where it measures.
MR. GROOME: Correct.
CHAIRMAN WALLIS: And then from that it
tries to predict the flow.
MR. GROOME: Right. In those -- in those
areas --
CHAIRMAN WALLIS: And if it has an S-shape
velocity profile, it's going to go --
MR. GROOME: It doesn't -- it doesn't know
how to calculate that, right.
CHAIRMAN WALLIS: But it still knows to
calculate the local velocity. That's what you --
MR. GROOME: Right. So it measures
essentially a velocity and it knows the area.
CHAIRMAN WALLIS: So you might actually
try looking at that velocity and recasting it in terms
of what's actually happening in the pipe and getting
a better sen- --
MR. REYES: And try to get the better,
yeah. Yeah.
MR. SCHROCK: Is it in the proximity of
the bend? Most everything is.
CHAIRMAN WALLIS: Everything is, yeah.
MR. REYES: How many -- how many l over
ds, John, from -- did we require for the --
MR. GROOME: It's -- l over ds, yeah, it's
right there on the bend. It's maybe like 2 d.
Actually I have a slide if you'd want to go back and
pull it up and we could actually look at a plan view
of it. But it's maybe like one and a half d from the
bend.
But that's fortunate for magnetic flow
meters to have the, you know, the least-required l
over d for flow measurement, or typically anywhere
from 1 to 5 d.
MR. REYES: Okay. So what we saw then was
loop stagnation due to the steam generator, one of the
steam generators becoming a heat source instead of
heat sink. But there was a delay, and we were curious
about that. So we investigated a little bit further.
And what we realized was that we're
injecting cold water to downcomer. And so we're
cooling off the downcomer. Of course, we have core
heat. And so we are creating a density difference in
the driving head. So the downcomer driving head,
because the system is completely liquid filled, is
still driving the flow.
And so what's happening is that the
reverse-heat transfer from the steam generator is
acting like a break. It's resisting the positive
flow, but it's not able to stop the flow. And so
that's why we see a continued natural circulation for
that period of time.
Eventually we get to a large enough delta
T on the steam generator where it is able to overcome
the driving head produced by the downcomer injection.
So you do have a downcomer recirculation.
That was -- that plays an important part
when you're working -- when you're looking at an
integral system injection. Because what that means is
even when you have "stagnant conditions," as soon as
you begin injecting, if you have any core heat, you
will create some natural circulation flow. And so you
-- by virtue of injecting you're producing a natural
circulation flow. So it's kind of --
CHAIRMAN WALLIS: As long as the injection
goes the right way.
MR. REYES: That's right, into the
downcomer. That's right.
Okay. So we observed those phenomena.
Steam generator tube voiding. That should
be the loop -- the cold-liquid plug --
CHAIRMAN WALLIS: If the steam generator
wins over the downcomer, then presumably the flow goes
the other way, if that's possible.
MR. REYES: The resistance is pretty large
in the other direction.
CHAIRMAN WALLIS: I was thinking of the
buoyancy. You're saying the buoyancy in the steam
generator counteracts the buoyancy in the downcomer.
MR. REYES: Right.
CHAIRMAN WALLIS: Presumably there's a
situation where the steam generator could win and the
flow could go the other way. Don't pull that cold
stuff into the loop seal, --
MR. REYES: Well, I think --
CHAIRMAN WALLIS: -- ready to come back
again.
MR. REYES: I think what -- of course, the
top of the plant will just get real hot and stay hot
and the cold water will stay on the bottom, so...
Okay. So we identified three modes of
loop stagnation, which we were interested in
understanding for integral tests.
One of them, the loop seal, the cold plug
in the loop seal actually is tied to a local
phenomenon, which I thought was one of the key
findings as far as relating our separate effects test
to the integral system test.
The other thing was the presence of this
RCP Weir wall. That really does delay the loop seal,
the formation of a cold-liquid plug in the loop seal.
And hence it delays stagnation in the loops.
So if we didn't have that Weir wall we
would have expected possibly to have stagnated a bit
earlier, because we would have formed cold plugs in
the loop seals. So that's kind of an interesting
result. So having them actually delays stagnation.
MR. SCHROCK: Isn't that Weir an attempt
at simulating the real plant?
MR. REYES: That's right, yeah.
MR. SCHROCK: And so how do you judge the
quality of that simulation?
MR. REYES: Right. Now at this point all
we can do is use our, in this case, CFD Codes to see
if we can, first of all, benchmark the CFD Code
against our test.
And then have a little bit of confidence
then that we can go forward and try to do a more
accurate model of the real plant.
Now in talking to the folks at Palisades,
they do indicate to us -- I mean they gave us the
dimension of this lip. And they have indicated to us
that, in fact, when they are draining, trying to drain
the cold legs they can't drain it completely because
of this lip. So we know it's an effect.
The other thing that it does cause -- that
we do observe is that the presence of that Weir wall
also, with side injection, we're going to see a lot of
stratification.
The plume doesn't come in from the top of
the pipe and then mix on the way down. So typically
the flow is going to be in the positive direction,
because it's not mixing on the way down, and you've
got this wall, this dam on the other side which, in
essence, is driving all the flow towards the downcomer
up to a certain point.
We said that reverse-heat transfer can
either -- it can reduce or stop the primary loop
natural circulation, depending on the available
downcomer fluid driving head.
Okay. So that's what we saw in the area
of loop stagnation.
CHAIRMAN WALLIS: What are your scaling
laws now? I would think that you have some scaling
laws for your loop.
MR. REYES: Correct.
CHAIRMAN WALLIS: And then I would think
that some sort of Froude numbers scale these, whether
it goes over the Weir and the stratification.
MR. REYES: That's right.
CHAIRMAN WALLIS: Is that consistent with
your scaling of other things like velocities and
dimensions?
MR. REYES: Yes.
CHAIRMAN WALLIS: It is?
MR. REYES: Yes. And so what we're doing
now is -- we have run a preliminary series. And we're
going to run several just to be sure. We're going to
identify the conditions for the onset of the loop seal
spillover, which is a very nice project for any
volunteering students.
It's a very straightforward effort. And
we know basically what the dimension of those groups
should be. And I think we can do a good job on that
one. So we're gathering that data right now.
MR. BAJOREK: Jose, your pump is down at
the bottom --
MR. REYES: Our pump is at the bottom of
the --
MR. BAJOREK: -- of the loop seal as
opposed to up at the top.
MR. REYES: Correct.
MR. BAJOREK: Are there any additional
restrictions in a regular PWR pump lower than the Weir
that a positive steam flow through the loop would
prevent some of that liquid coming back over the Weir
and delaying the cooling?
MR. REYES: Now we're talking steam flow?
MR. BAJOREK: Yes.
MR. REYES: There was a schematic that
John had showed us yesterday, and it's a fairly short
-- what it is it's kind of a relatively flat --
MR. BAJOREK: Impeller.
MR. REYES: -- impeller with -- in a
volute. Okay. I was looking for the right word, a
volute. So if that comes -- your cold leg comes out
at this angle, and then you've got this -- a short
drop and then to the loop seal.
So I don't think there's anything other
than the impeller itself that would hinder the steam
from getting up there and going out. Is that -- I'm
not sure if I understand.
So you're going to have -- you're going to
have -- the steam will come in, and if there's water
on the bottom of the cold leg, it's basically just a
stagnant pool almost, except for right at the surface.
Okay. Now we'll talk a little bit about
-- if there's no other questions on stagnation.
Again, the reason we looked at stagnation so heavily
was because in previous studies the thought was, well,
if the primary -- if the cold legs are stagnant, then
injection under those conditions would essentially be
the worst plume conditions. Okay. As we continued
with our study we saw that we came to a different
conclusion.
COLD LEG THERMAL STRATIFICATION AND
PLUME FORMATION IN APEX-CE
MR. REYES: So I'll talk a little bit
about cold leg thermal stratification first, and then
we'll talk about the plume behavior in the downcomer.
Okay. We did these flow visualization
tests, which were really very helpful. We started by
just doing a series of tests in APEX at pressure and
temperature. And we were measuring our -- we had our
thermocouple rake in there.
We were seeing constantly that we had cold
temperatures at the bottom for most of the flow rates
that we generated, the cold-leg flow rates. And we
were curious then what was going on. So we built this
very simple flow visualization test that allows us to
take a look at the similar Froude number conditions,
what might be going on in the pipe. And we --
CHAIRMAN WALLIS: This is green salty
water; is that what it is?
MR. REYES: That's green salty water. So
we're injecting -- we're using sodium fluorescein to
actually give us that green color. And if you hit it
with ultraviolet light you get a very bright image
with that.
So we've put sodium fluorescein in our
salty water. And we're using that to represent our
cold injection.
The pipe is initially filled with fresh
water. And that whole tank actually is filled with
fresh water initially.
So we begin our injection, and this is the
type of behavior that we see.
CHAIRMAN WALLIS: Where is the injection
here?
MR. REYES: I lost my mouse. Oh, here it
is. Thank you. So --
CHAIRMAN WALLIS: It's the pipe that we
can't see, which is there?
MR. REYES: Yeah, that's right. It's a
side injection. And so from a top view the injection
line would be behind the pipe.
CHAIRMAN WALLIS: That's why we can't see
it.
MR. REYES: That's right. That's right.
So... Thank you.
CHAIRMAN WALLIS: So is this fog flowing
down the Columbia Gorge.
MR. REYES: It's the fog coming down the
Columbia and out towards the ocean here, I guess.
Here's the Weir wall. So for this
injection flow rate and for this combination of
cold-leg flow rate, we're not getting it -- well,
maybe we are spilling over a little bit here. Here's
a side view. So we're not, okay. So for this --
CHAIRMAN WALLIS: It doesn't look like a
very flat layer, does it?
MR. REYES: Say again.
CHAIRMAN WALLIS: It doesn't look like a
very flat interface between the green and the blank.
MR. REYES: It's not very flat. For this
flow rate it's not very flat. That's right. So what
we're looking at then is a -- we're at about the 12-
gallons per minute in the cold leg. And we're
injecting probably somewhere around a gallon per
minute here through the -- so you see the side
injection.
The plume comes in. It travels both
directions, and back towards the vessel and towards
the loop seal. But this acts as an effective wall to
prevent the flow.
CHAIRMAN WALLIS: I guess the drag from
the other flow keeps the -- it's what causes it to
flow in this arrangement.
MR. REYES: So here's a closer look at it,
and I'll show you the Weir wall.
But here we're injecting again. This is
-- you can see this --
CHAIRMAN WALLIS: Well, that shows how
stratification inhibits the mixing.
MR. REYES: Right. Here's a close-up of
the -- of that --
CHAIRMAN WALLIS: You can't just use the
normal CFD in a stratified flow like that. K-epsilon
doesn't work in their stratified interface.
MR. REYES: That's one of the -- this is
where -- this is -- Ian's nodding his head yeah.
Working with CFD Codes, and we'll talk
about that a little bit later, but that's some of the
challenges that we face with using these types of
codes. And so we're going to be looking for some
advice from experts.
CHAIRMAN WALLIS: The turbulence is really
suppressed at the interface.
MR. REYES: Yeah. So this is at the Weir
wall, and we see that basically the cold fluid from
the cold leg -- I mean the cold-leg fluid is sweeping
all that fluid back.
Okay, another look here. We have a little
bit of spillover in this case. We've got the flow
rate. And then this is looking again at the spillover
into the loop seal. So we'll perform some of those
visualizations for you when we go over there after
lunch.
Okay. So let's talk a little bit about
what we saw with regards to the stratification in the
cold legs, some of the measurements from APEX-CE.
So we did --
CHAIRMAN WALLIS: Now are you going to
tell us about analysis of this in terms of some math
and some predictions?
MR. REYES: Yes.
CHAIRMAN WALLIS: Somebody -- somebody is.
MR. REYES: When we talk about the plumes.
CHAIRMAN WALLIS: Somebody is. Well, what
about the cold-leg behavior and the stratification and
all that?
MR. REYES: We have some predictions to
show you later on if you -- we can chart --
CHAIRMAN WALLIS: Any comparisons with
data?
MR. REYES: We have lots of data.
CHAIRMAN WALLIS: Do you have -- well,
come on.
MR. REYES: We'll give you some. You'd
like some theoretical...
Well, we have looked at the plumes in the
downcomer, and I'll present some equations there.
In the stratified region, we're still
looking at developing what's going on in that section.
We have a -- there was a -- I think I know
what you're referring to. Early on in the development
of this project we were real curious about the onset
of thermal stratification and came up with pretty good
criteria that was used -- that you could use to
predict when you have essentially well-mixed
conditions in the cold leg.
Now the presence of this Weir wall has
changed that somewhat. So we're looking -- that's why
we're changing the theory some, to examine that Weir
wall more closely and understand how that's affecting
the flow and what that really means in terms of a
stratification criteria.
CHAIRMAN WALLIS: It looks like a real
candidate for the Kelvin-Helmholtz instability type of
analysis.
MR. REYES: That's right. That's right.
So now you've got -- I mean that's exactly right. So
that's what we're looking at right now.
So what this tells me then is that the
stratification criteria that we use that P. F. Foss
had developed, and then the modified version which I
had developed, which was based on the Froude number of
the cold-leg stream squared plus the Froude number of
the hot leg squared equal to one.
With the presence of that Weir wall we
need to look at that and say, okay, now we've got a
slightly different situation.
CHAIRMAN WALLIS: What is 10 gpm in terms
of Froude number, using the density difference.
MR. REYES: Oh, 10, that -- with using
that combination, I think it's like .04. That's that
modified Froude number. That --
CHAIRMAN WALLIS: You're using the density
difference between the fluids?
MR. REYES: Density difference between the
fluids. And using --
CHAIRMAN WALLIS: That's the only Froude
number. It's not modified. That is the Froude
number.
MR. REYES: Well, this -- they're modified
in terms of -- it uses the HPI injection.
CHAIRMAN WALLIS: Density.
MR. REYES: Density --
CHAIRMAN WALLIS: Minus --
MR. REYES: -- with the cold leg --
CHAIRMAN WALLIS: Minus the density of the
--
MR. REYES: Oh, okay, I'm sorry. The
Froude for the cold leg. Yeah, I'll have to look that
one up. I was giving you the injection, injection
Froude. Yeah, I can look that up for you.
CHAIRMAN WALLIS: You can tell us after
the break.
MR. REYES: You bet. We've got lots of
calculators.
So what we found, though, is that the
presence of the Weir wall results in some
stratification for the full range of conditions that
we studied.
So this test 003, what we were doing was
we were putting -- we were parametrically varying the
cold-leg flow rate and the injection flow rate for 16
different cases. We wanted to see what we would
observe as far as stratification in the cold leg.
So what we saw was that at the presence of
the Weir wall there was always some stratification for
the flow conditions that we looked at. And we were
looking at essentially from one and a half percent to
K to four percent to K powers over a range of about
30- to 100-percent HPSI injection flow rates.
The spillover was not observed in any of
these tests, which meant that our -- above 30-percent
HPSI for us, what we saw was that we had enough flow.
It was greater than 10 gallons per minute in the --
excuse me -- for one-and-a-half percent to K power,
our flow in the cold legs was greater than 10 gallons
per minute.
So we always kept flowing in the direction
of the reactor vessel. And there was no spillover for
any of these tests. So this is the range of
conditions we cited.
We would vary the K power. That would
change our natural circulation of flow rate in the
loop. And in between each test we would turn our
pumps for a while and get everything back to uniform
temperature and then do another parametric study.
So here's that --
CHAIRMAN WALLIS: What's the basis of
these HPSI flow rates? What's the basis for choosing
these values?
MR. REYES: Oh, that's -- that basically
was the limits of what we could do. Yeah. So 30
percent, when you drop below 30 percent on our
injection, we have difficulty controlling our -- well,
for -- for the cases that we looked at, I guess we can
get down to --
CHAIRMAN WALLIS: Presumably it's a scaled
HPSI from reactor conditions.
MR. REYES: Right. Right. So point --
the lowest we did was .35 gallons per minute, which
corresponds to about 30 percent of one injection flow.
So that was what --
CHAIRMAN WALLIS: So this is a throttled
HPSI of --
MR. REYES: Right.
This shows the range of tests that we did.
And in between each test, again, we operated our
reactor coolant pumps.
Once we put in several hundred gallons per
minute through the loops. Of course, everything warms
up pretty uniformly. All our thermocouples matched
up.
This is looking at the top of the cold leg
number 3 and this is the bottom of cold leg number 3,
so we're going -- we're looking at the temperature
stratification in the cold legs.
And this shows that for the different
conditions we saw we always saw some stratification.
Even at our minimum being .35 gallons per minute, we
saw some stratification in cold leg number 3.
Again, this being the bottom of the cold
leg here and all the way up at the top would be the
top, so they actually went in order. So we saw that
stratification for our tests.
Here we are at -- for cold leg number 4,
the same test, we were running -- we ran two different
injection flow rates for that --
CHAIRMAN WALLIS: What is your temperature
of your HPSI?
MR. REYES: HPSI temperature is about 65
degrees Fahrenheit.
CHAIRMAN WALLIS: So this is a lot hotter
than the HPSI itself.
MR. REYES: Oh, much, much hotter, right.
Yeah. So what we're seeing is by the time we get to
this -- the typical rates, we are seeing some -- quite
a bit of mixing.
And the total -- well, I mean this is 150
degree delta T from the bottom of the pipe to the top
of the pipe, so it's a pretty big stratification.
This case and this case, we also used to
model with STAR-CD. So these two cases we modeled
with STAR-CD. And later on you'll see the results of
those comparisons.
CHAIRMAN WALLIS: So maybe if you had
bottom injection you might actually have those minima
going down to something like the HPSI temperature.
Because presumably it's bottom injection; it doesn't
mix with anything.
MR. REYES: That's right.
Now I'm not sure if there are any plants
that do bottom --
CHAIRMAN WALLIS: I don't think there are,
but just for comparison sake.
MR. REYES: -- with the big --
CHAIRMAN WALLIS: I mean you said with top
injection there's more mixing and side injection you
get more of this.
MR. REYES: Right. So the bottom might be
--
CHAIRMAN WALLIS: Presumably there's one
limit where you just ooze the cold water in and it
flows along without mixing with anything.
MR. WACHS: You get conducted heating from
the metal in the cold leg.
MR. REYES: Dan.
MR. WACHS: I said you'll still get
conducted heating from the metal in the cold leg, so
there will still be some warming. You won't -- you're
unlikely to --
CHAIRMAN WALLIS: Yes. Yes.
MR. WACHS: -- get that small.
CHAIRMAN WALLIS: Yeah, I guess you need
to estimate that, too.
MR. REYES: Okay. So the upshot of it for
us for these series of tests was that for all the
natural circulation cases that we examined, we always
saw some stratification.
And this case here being the maximum
stratification we observed, which was not the -- not
the lowest flow rate, but it was -- it was essentially
close to the highest flow rate, but not the -- not
necessarily the lowest. Well, like big yeah, the
trend is getting bigger as we go to lower -- lower
cold-leg flows.
Okay. So we always see some
stratification, which was different than what we've
seen in the past. If we kept the criteria, the
stratification criteria the way it was, it would
predict good mixing for some of these tests.
So we see that the Weir wall has no -- has
an effect. And we need to change that theoretical
model.
Okay. Now we'll talk a little bit what's
going on in the downcomer as far as plumes.
I'll start off with kind of a typical
analysis and what's been done in the past. The
classic analysis is you have a single planar plume
falling into a stagnant, uniform ambient fluid.
Okay. That would be the classic analysis.
And it's been done -- it's been done for a long time.
Bachelor did a study on it and Morton and Rouse and --
CHAIRMAN WALLIS: You say, "planar." You
mean it's 2 d?
MR. REYES: Correct. Yeah, they're 2 d.
CHAIRMAN WALLIS: They're also
cylindrically symmetrical. The simple ones are --
MR. REYES: Right. The --
CHAIRMAN WALLIS: -- planar or --
MR. REYES: Right, the axisymmetric --
CHAIRMAN WALLIS: -- axisymmetric.
MR. REYES: Right, axisymmetric case. You
just change the coordinates and unwrap it and get a
planar. Yeah, so the 2 d case.
And these involve some very classic
assumptions, some very -- which work very well for the
single planar plumes and also for the axisymmetric
plumes.
CHAIRMAN WALLIS: Oh, review that for me.
If I have, say, a faucet, the plume actually
accelerates instead of -- because the density
difference is so enormous.
MR. REYES: Right. Yeah.
CHAIRMAN WALLIS: And then if I have a
very low-density difference, presumably the buoyancy
is not so big and the plume spreads a lot. What sort
of range are you in in terms of the density
difference, in terms of, you know, whether the plume
spreads a lot or --
MR. REYES: Right.
CHAIRMAN WALLIS: -- doesn't because of
the gravity accelerating it?
MR. REYES: Right. For these tests, our
delta rho over rho was like .18. So I'd have to go
back and see what just a delta rho is. But -- so it's
comparable to what you would see in the plant, not --
not as large, but it's within 10 percent.
CHAIRMAN WALLIS: But it's something like
the plume from a cigarette, or something, in terms of
what you --
MR. REYES: Oh, no.
CHAIRMAN WALLIS: -- imagine in terms of
things you know about?
MR. REYES: Okay. So smoke and air, maybe
-- maybe like --
CHAIRMAN WALLIS: Smoke from a chimney,
from -- on a clear day. Something like that.
MR. REYES: Yeah. I'm thinking of my
backyard.
CHAIRMAN WALLIS: Yes, something like
that.
MR. REYES: I'm looking at the stack way
out there.
CHAIRMAN WALLIS: But, you see, if you
have a hot enough fire --
MR. REYES: I'm looking at the stack way
down there.
CHAIRMAN WALLIS: -- your plume actually
--
MR. REYES: Yeah, it would be similar --
CHAIRMAN WALLIS: -- can go up and
concentrate.
MR. REYES: Right. So --
CHAIRMAN WALLIS: Before it spreads.
MR. REYES: -- you'll see a lot of
examples of the reverse. You'll see a hot plume --
CHAIRMAN WALLIS: Right.
MR. REYES: -- going up in air.
CHAIRMAN WALLIS: Depends on how hot it
is.
MR. REYES: So it -- from the shape of it,
I'd say --
CHAIRMAN WALLIS: So I'm trying to get a
feel for which kind of a plume is it. I think it's a
spreading plume.
MR. REYES: It's a spreading plume.
CHAIRMAN WALLIS: Yeah. You have the
fire, the density difference is two to one or
something. So it's...
MR. SCHROCK: These theories pertain to,
as you've said, a large field. You've got the
downcomer walls that are confining the plume. Are you
going to address that?
MR. REYES: I'll cite the difficulties
with those things, right. Yeah.
We'll go on, and I'll show you some of the
problems we're facing --
CHAIRMAN WALLIS: There's a good --
there's a good book on this by some German whose name
I forget, who studied all kinds of plumes and all the
--
MR. REYES: I've got it in my bag.
CHAIRMAN WALLIS: Okay.
MR. REYES: It's Rodi and Chatney.
CHAIRMAN WALLIS: Rodi, that's right.
Rodi.
MR. REYES: They did a ton of work in --
you're welcome to look at the book. Yeah.
CHAIRMAN WALLIS: It's pretty
comprehensive.
MR. REYES: It is. It was a very nice --
one of the few that covered a wide range of
axisymmetric and -- but I'll show you another paper
today, which is fairly new, which is closer to our
situation.
CHAIRMAN WALLIS: It's got too many ns.
MR. REYES: So we're familiar with the
classic assumptions for the planar plumes. The idea
that there's a linear spread of plume radius with
axial position, because you have a constant
entrainment rate. And that assumption works
reasonably well for the planar plumes in stagnant
media.
You've used the similarity of velocity and
buoyancy profiles. And you can come up with some
universal curves that way. And in REMIX they kind of
use that technique of producing these universal
curves. And then at different locations it says,
okay, for this set of dimension groups you kind of
convert it back into what you should be reading there,
so that works well.
And, of course, they always -- you
typically assume the Gaussian-shaped profile for this
stagnant media. I'll show you a couple of pictures.
Back in 1934, we've got some data which
shows the same situation. Here you have the velocity
measurements inside the plume. The scale's missing
here. That should be centimeters per second on the y
axis.
The velocity measurements for the plumes
at different axial locations. And if you -- you can
actually scale it with a mean velocity of plume and
come up with a single shape, a single universal
Gaussian curve. So we know that that theory works
very well.
So the thought was, well, we can apply to
some of the similar concepts to our test --
CHAIRMAN WALLIS: Well, I'm trying to
think, though. When you pour -- pull this stuff over
the lip of a pipe, --
MR. REYES: Right.
CHAIRMAN WALLIS: -- you don't have the
sort of starting condition of a uniform velocity. It
has to sort of accelerate out of the pipe. So it's
accelerating for a little while before it does this
mixing.
MR. REYES: That's right.
CHAIRMAN WALLIS: Isn't it?
MR. REYES: So -- yeah, I was --
CHAIRMAN WALLIS: I don't quite know how
you modeled the starting condition for coming out of
the pipe and going over the lip and into the
downcomer. How do you start your plume for an
analysis like this?
MR. REYES: For an analysis like this what
we think is actually happening is that the plume is
jumping the gap. Eventually it's --
CHAIRMAN WALLIS: It hits the other wall.
MR. REYES: -- it hits the core barrel
wall. Now you've got another, another problem. So
that's what we believe happened.
CHAIRMAN WALLIS: This is like the
experiments that we did with the injection of water
into a steam down- -- filled downcomer. It jumps to
the wall and goes down the other side.
MR. REYES: Which test was that?
CHAIRMAN WALLIS: When was that done?
MR. REYES: Yeah.
CHAIRMAN WALLIS: 7-70, or something like
that. 1970 maybe.
MR. REYES: Yeah. For our test number 13
we're thinking of something similar, so I'll ask you
a little bit more about that.
The -- so that's right. There's some type
of flow establishment region. If you have a forced
flow, you have a momentum dominated, then some kind of
a transition. And then you eventually gets to this
buoyancy-dominated region for the plume.
So this region actually could be fairly
short. And where you'd like to be is kind of in this
region as far as analysis. So you've got this
spreading plume. But again that's for a stagnant
medium.
If you have -- if you have any
complications to it, you really have to -- I mean it
complicates the analysis quite a bit. If you're
impinging onto a core barrel wall, well, how do you
analyze that.
So we very quickly got -- as we started
looking at the different complications, that's where
we said, well, we need to use to some of the CFD and
see if we could understand that a little bit better.
CHAIRMAN WALLIS: Now let's look at Mr.
Foerthamnn's experiment.
MR. REYES: Sure. Can you go back?
CHAIRMAN WALLIS: Now it seems as if the
width of the plume is about four centimeters, or
something, but the velocity in the middle hasn't died
to half the initial one until it's gone 75
centimeters. So this is a plume that's very
persistent. It's gone an awful long way before its
velocity in the middle has gone down by a half.
MR. REYES: Right.
CHAIRMAN WALLIS: So your plumes aren't
anything like that. Yours --
MR. REYES: No.
CHAIRMAN WALLIS: -- spread much more
rapidly, don't they?
MR. REYES: Correct. Correct.
CHAIRMAN WALLIS: Why is that? There's
something different about some dimensionless number in
your experiment than in this one?
MR. REYES: Right. Now for this
experiment --
CHAIRMAN WALLIS: There must be something.
MR. REYES: I think -- yeah. There -- I'm
not exactly familiar with the delta rho over rho in
this test or the density difference in this test and
what was going on as far as the buoyancy.
CHAIRMAN WALLIS: I think this is actually
spreading less than the jet would in just -- the
shrifting-type jet with no buoyancy at all. This one
is actually buoyant, so it's spreading less than a
jet. And --
MR. REYES: Than the actual.
CHAIRMAN WALLIS: -- that's why it's so
surprising that your plumes spread so rapidly.
MR. REYES: Well, there's other -- so
there's other -- right. So there has to be other
mechanisms that are --
CHAIRMAN WALLIS: You think about the
plume from the stack from the incinerator next to your
-- do you have an incinerator next to your building?
MR. REYES: It's a -- well, I can see it
from --
CHAIRMAN WALLIS: That probably goes quite
a long way on a clear day before it spreads much.
MR. REYES: That's right.
So what we -- we started looking at our
plumes and realized that there's something -- it's
significantly more complicated, especially for the
conditions that we were looking at, because we had
multiple asymmetric plume interactions. We actually
had some cases where we had cold flow.
So now instead of a stagnant downcomer,
what you actually have is hot-leg flow. You're
putting hot water -- the way this was stratified,
you're -- basically the cold water is pouring --
pouring out the bottom of the pipe.
You still had positive hot-water flow on
the top of the cold leg going into the downcomer. And
you were forcing this flow through the downcomer co-
current with the plume. So now you've got a
could-current situation.
CHAIRMAN WALLIS: That could be worse.
MR. REYES: Yeah. So now I'm starting
thinking relevant velocity between the plume and your
medium. If you have a stagnant case, your plume comes
in and the relative velocity for that case is going to
be larger than if the media traveling with the plume
is at nearly the same velocity. So now the relative
velocity between the plume -- and that's a well
documented -- that's this plume behavior in cold flow.
And so we searched the literature for
that, and we found a paper that was fairly recent that
tried to address that, by Wood.
So potentially with some cold flow you can
actually have less -- you'll preserve the plume
further under certain conditions, okay.
There are other compounding conditions, of
course. We -- and our tests, of course, as opposed to
some of the studies done previously, we have a core
with a core barrel. And that's dumping heat into the
downcomer. So there are other things which are
heating up this plume and causing it to dissipate a
bit earlier.
So you have these -- all these different
factors, some trying to preserve the plume and some
trying to destroy the plume. Okay.
So we very quickly came to realize that if
we wanted to do a realistic or something realistic,
that maybe we can bound the problem asymptotically and
look at some of the ranges of the plume spreading.
But, in fact, the thought then came, we
need to do some type of CFD work. And that, of
course, brought with it its own set of problems, how
to understand CFD and who would volunteer to do the
work.
And so we'll present that later on. And
we're wide open to suggestions on that. But you'll
see some of the comparisons, and you can be the judge
of what the problems might be or what went right.
Okay. So, in essence, there is a test
facility in Finland, Imatran Voima Oy, which in the
good old days they did these flow visualizations. And
I guess they had an artist watching the stuff, because
this is an artist sketching. He must have been very
fast.
CHAIRMAN WALLIS: Well, Leonardo did it
all 500 and some odd years ago.
MR. REYES: That's right. I think this
may be a da Vinci.
CHAIRMAN WALLIS: But Creare did something
like this, too, didn't he?
MR. REYES: Yeah. So Creare had that --
well, they only had the single.
CHAIRMAN WALLIS: It was red or blue, that
stuff that they used, but it was -- they had pictures
like this, I think.
MR. REYES: They did? I don't think -- I
think they had a single injection, a single --
CHAIRMAN WALLIS: Unwrapped?
MR. REYES: -- cold leg.
CHAIRMAN WALLIS: I thought they had an
unwrapped --
MR. REYES: Well, maybe they did have two.
CHAIRMAN WALLIS: Maybe -- well, go back
and find out.
MR. REYES: I don't know if they ever --
I never saw they had for two. Yeah, I saw two tests
that they did.
But they had -- they were using that red
dye and they were injecting into each of the cold legs
and sometimes they flow, but they were getting some --
they were getting plume merging. And so, of course,
that -- you know, what's the strength of the plume
where they -- when they merge and how does that affect
heat-transfer coefficient, da-da-da. It goes on and
on.
So in terms of simple analyses, they very
quickly -- it became obvious that it was beyond what
you can do on the back of an envelope --
CHAIRMAN WALLIS: That plume seems to be
going a lot more than 4 ds. You were talking about
mixing by 5 ds. Or the beltline is at 5 ds from the
pipe, or something?
MR. REYES: Right. For Palisades it's
somewhere up there --
CHAIRMAN WALLIS: And this one seems to be
going down a lot more than that --
MR. REYES: This -- this -- yeah, this
looks like it's going down for -- now the flow rates
for the -- for what they were doing were very, very
high in these tests. So that's a big part of it.
And I think we have a picture later on
that we'll show you that kind of goes back to this
test. And, of course, we saw similar behavior as far
as merging, but we didn't see the plumes, you know,
getting --
MR. BESSETTE: You have to remember also
this is an artist's rendition.
MR. REYES: So da Vinci --
MR. BESSETTE: Done by -- Actually it was
done by Tuemisto himself.
MR. REYES: Tuemisto, okay.
MR. BESSETTE: Yeah.
MR. REYES: Tuemisto da Vinci.
CHAIRMAN WALLIS: They didn't have
photography in those days.
MR. REYES: We have some --
MR. BESSETTE: Not -- well, I don't know.
Like I say, he'd interpreted what he saw, and he did
it by drawing.
MR. REYES: So when we started looking at
the plumes we were having difficulty interpreting what
was going on in the downcomer. So Kent Abel, one of
our graduate students, came up with this idea of
putting together this, an unwrapped map.
And so each -- this shows all four of the
cold legs. On top of it, it gives you the flow rate
through the cold leg, and then the HPSI flow rate for
each of the cold legs. So those are listed up on top.
And then along side here he's got a color code for the
different temperatures.
And then here in the cold leg you can see
if the cold leg is stratified or not, so again by
temperature. So we can observe stratification. When
this light is green, it means that the HPSI is flowing
for that particular transit.
You can pull up any one of the tests that
we've performed into this format, and it'll just play
the downcomer for you, which I love it. I can -- I
sit there and watch the lights.
CHAIRMAN WALLIS: Do you have music?
MR. REYES: HPSI, HPSI. You get
hypnotized kind of by it.
But this is just an example. If we have
time I think we can run one of these or two to show
you a couple of different scenarios and what we see.
It's useful because we break it up into a
gradient, a delta T that your eye can actually catch
instead of a continuous type of a thing.
And what we see in this particular
snapshot, we see this plume, we see cold temperature
over here and cold temperature over here, so we're in
the dark, this dark blue region over here.
CHAIRMAN WALLIS: It's not very obvious.
MR. REYES: It's not real obvious, but it
does interact --
CHAIRMAN WALLIS: Can't you get more
contrast?
MR. REYES: Right. In some of the other
-- I think in some of the others we'll see more
contrast.
CHAIRMAN WALLIS: Yeah.
MR. REYES: But the thing is that the
temperature difference is not -- you're within --
they're typically within 10 degrees or so, 10 or 15
degrees. And so we could do some more contrast.
So what we see is a kind of interaction
here, and then we do see colder here and some cold
down here. So we kind of see a merging plume over
here. But then again the ambient is relatively cold
also.
So in terms of a delta rho, we're not
looking at a very large delta rho. However, we are
seeing this merging and then getting down blow here.
So you can imagine that if we had a couple thousand
more thermocouples, that would have been ideal.
So that's one example.
Now you can see it a little bit better as
we run -- is that the next -- yeah. We'll run this
for you, and then you can see a little bit better the
temperature changes. And this was for the -- this was
for the main steam line break case, so you'll see
eventually everything will turn blue because the --
CHAIRMAN WALLIS: Yeah. I think if it's
dynamic it will be easier to see.
MR. REYES: Right. And I think that's
what we'll do.
Okay. So it's running right now. We're
getting -- the time on the upper left-hand corner is
the time to start the test. And so it's jumping.
We're jumping about -- every step is about eight
seconds or so.
And this is the main steam line break. So
we've opened up the break, and so you see all the
temperatures are uniform. Now the break's open.
We're starting to see cooling, but that's not cooling
to the HPSI.
Now the green -- the HPSI's on now at this
point, and so you'll start seeing colder temperatures
underneath the cold legs. So there you go. So you
see some of the yellow and the green, so it's getting
colder underneath there.
CHAIRMAN WALLIS: Can you -- uh-huh, okay.
MR. REYES: And we can step through it if
you want to go slower. So you see --
CHAIRMAN WALLIS: So this is typically the
whole downcomer, so the beltline is sort of where in
this, in the middle --
MR. REYES: The beltline is between the 4
d and the 8 d.
CHAIRMAN WALLIS: Somewhere in the middle
of the page.
MR. REYES: Yeah. And we -- and that's
actually where our beltline is. We have a big flange
sitting there so we couldn't get thermocouples in it.
And so you can see at this point the cold
leg number 3 there is still so much stratified. Cold
leg number 4 is completely cold, but then that was
where the broken steam generator was. So you're
seeing some cold flows there.
So now that the whole downcomer is
overwhelmed basically by the transient itself, the
steam line break. Looking --
MR. BOEHNERT: So that was every eight
seconds, was it?
MR. REYES: Correct. Correct. Yeah, I
think we're -- were we jumping eight seconds there,
Ken? Yeah.
CHAIRMAN WALLIS: If you go back to about
one or two seconds after it began to turn yellow, --
MR. REYES: Okay.
CHAIRMAN WALLIS: -- you've got some --
some spludges pretty low down of the -- a really
different color from the surroundings. I got the
impression that there was some plume activity down to
maybe 10 d or something. Not insignificant.
MR. REYES: Right. Well, -- yeah. In
fact, we'll talk about what happens when you have that
cold-leg flow and you're trying to preserve the plume.
We see it -- the behavior is somewhat different. I
won't steal -- I won't steal their thunder.
But, yeah, we do see a different behavior
for cases where your cold legs are flowing and -- for
this plant.
(Discussion held away from the microphone and
simultaneous talking.)
MR. HAN: One of the things to keep in
mind --
MR. REYES: It was one of our
thermocouples --
MR. HAN: -- in terms of the vessel is
that transient --
(Discussion held away from the microphone.)
MR. HAN: -- behavior is not important
because of the time constant of the vessel wall --
MR. REYES: This is James Han.
MR. HAN: -- things that happened over the
time of, say, one minute, for example, don't matter
because you don't build up a stress. So you can have
these plumes. Let's say, if they're moving and
whatever, that the transient behavior doesn't matter.
It's the longer-term behavior that's important, things
that happen over the course of 15 minutes to 45
minutes.
CHAIRMAN WALLIS: It's a question of the
balance between the surface h and what you think of
sort of internal heat transfer distance of the wall.
MR. BESSETTE: And it's conduction -- it's
conduction --
CHAIRMAN WALLIS: It's certainly not an
infinite h on the surface.
MR. BESSETTE: No, but if it does -- the
h doesn't matter. It's because it's so conduction
controlled.
CHAIRMAN WALLIS: Yeah. So it's like an
infinite h?
MR. BESSETTE: It's like an infinite h.
If you had an infinite h or not an infinite h, it
doesn't really matter too much.
CHAIRMAN WALLIS: So you're immediately
chilling the surface to the temperature of the water?
MR. BESSETTE: Essentially your boundary
layer doesn't matter that much. It's the ambient --
CHAIRMAN WALLIS: But you had to know h
because if you assumed h is infinite it's not very
nice.
MR. BESSETTE: Well, --
CHAIRMAN WALLIS: You have to know h.
MR. BESSETTE: -- if H is infinite -- if
h is infinite, it means you have no boundary layer.
It means your ambient temperature is the temperature
of the surface of the wall.
CHAIRMAN WALLIS: I don't think you want
that.
MR. BESSETTE: It doesn't matter that
much.
CHAIRMAN WALLIS: It stresses the wall.
That temperature difference is what stresses the wall,
the difference from surface from the average, is it?
MR. REYES: There is a thermal penetration
of time that --
MR. BESSETTE: The difference between the
ambient fluid temperature and the wall surface
temperature is never large, no matter what h is.
CHAIRMAN WALLIS: Then you might as well
assume h is infinite for your -- and forget about
everything else as long as you know what the
temperature is.
MR. BESSETTE: That's right. You can do
that.
CHAIRMAN WALLIS: I don't think that's
very good for -- well, maybe we -- that's a different
discussion somewhere else.
I thought it was pretty critical what h
was.
MR. BESSETTE: No. In fact, we've done
those sensitivity studies already --
CHAIRMAN WALLIS: Okay. So we'll go back
to that some time.
MR. BESSETTE: -- to show that h is not
important.
MR. SHACK: I mean that's good news.
MR. BESSETTE: Yeah.
MR. HAN: That's good news, yeah.
CHAIRMAN WALLIS: So we just forget about
thermal hydraulics.
MR. SHACK: No, no. You had to set the
temperature.
MR. KRESS: No, you need the temperature.
CHAIRMAN WALLIS: Okay.
MR. BESSETTE: We've been trying to tell
you that.
(Laughter.)
MR. SHACK: But it's still your
penetration depth that you're interested in, isn't --
CHAIRMAN WALLIS: See, there's a white --
there's a light-colored one way down there, right?
Presumably that -- what's happening there? It looks
very strange to me. You've got colder stuff down
there than you've got --
MR. KRESS: At the middle it curled over.
CHAIRMAN WALLIS: Yeah. But it's just
going to warm again at the top, isn't it? Or, no,
maybe I've -- yeah, it's going to warm again at the
top. There's a plug, sort of a lump of cold fluid in
there not even connected to the pipe.
MR. REYES: Yeah. And sometimes what you
see is -- well, our thermocouples -- that's a good --
that's a good point.
Our thermocouples are closer to the vessel
wall because that's where we wanted to measure, what
was impacting the wall.
CHAIRMAN WALLIS: Oh, so if it jumps away
to the inside, --
MR. REYES: If it jumps --
CHAIRMAN WALLIS: -- you won't see it?
MR. REYES: Yeah. So quite often --
that's right. So quite often what we see is when
we're injecting -- it'll jump that first thermocouple.
And that one will read hot. And all the thermocouples
below will read cold. And so it's impacting the
barrel wall, mixing up, and then it's --
CHAIRMAN WALLIS: But here we have -- do
those yellow things -- are they those things -- they
look -- that doesn't look like a plume. It just looks
like a -- sort of a odd-shaped lump of fluid.
MR. REYES: It's not like the -- well,
again we don't have -- we don't have a lot of
thermocouples to get a good fine detail of this thing.
But you're right, the shapes are not going
to be nicely-spreading plumes. There are going to be
some meandering --
MR. SHACK: Will we see this in the CFD
calculation, something like this?
MR. REYES: Well, yeah, I think you'll see
that the CFD calculation does more of the meandering
and the curling-type behavior, so...
Okay. I'll let them get us back to the
presentation.
Okay. So one of the things that we've
observed is that at least as far as -- in terms of an
asymptotic solution or trying to imagine a little bit
of what's going on, there was a paper recently by
Wood. It was called, "Asymptotic Solutions and the
Behavior of Outfall Plumes."
And it was very nice -- nicely done
because you don't -- it was the only one I found that
was fairly recent that talked about the spreading of
the plume under some unusual conditions, either
crossflow or co-flow.
So what is done basically is the b
represents the half width of the plume. Okay, and z
is the axial position of the plume as it comes down.
And so this basically describes the
spreading of the plume. So for a stagnant case, u
subinfinite here, that's a stream velocity. Up is the
plume centerline velocity. And ks is the spread
constant.
And so for a stagnant media where you're
injecting these plumes, k is a constant. I mean this
goes to zero. The ups cancel in case. So he was able
to kind of unify that behavior.
Whenever you have a flow, this is the
angle between the trajectory of the plume versus your
flow. So if it's co-flow, that would be -- the
cosine's zero is one. So you'd have up over up plus
u infinite.
So what this is telling us then is it's a
factor that makes this constant smaller and as a
result you get a plume width which is smaller, tighter
as you go down in your axial position.
So this confirmed an idea that, well, for
co-flow you would expect to see tighter plumes. They
might be preserved a little bit longer than you'd
expect to see in a stagnant media.
So this suggests that for certain -- so
now a scenario comes to mind. You have stagnant
conditions in your -- in your downcomer. You inject
the plume.
You'll get very good mixing because the
relative velocity between the plume and the ambient is
going to be higher than the case when you have co-flow
at low-flow rates. When you've got a co-flowing
plume, so you'd expect the plume to be preserved a
little longer. And we'll talk about the phenomena
that we see in co-flow versus stagnant because we did
both cases.
Of course, if your stream-flow rate
continues to increase, eventually something's going to
-- this model wouldn't apply. So there's got to be a
limit to this. And so that would be on the asymptotic
end of this study.
So if you'd get enough crossflow or enough
downflow, you'd expect the plume to break up because
of it. So again it's probably -- you've got a
relative velocity criteria and -- if I come up with
something that works reasonably well.
So again this just explains what I just
said, that the spreading for the stagnant condition,
you expect it to be greater than spreading for a co-
flow.
Now this is -- again this is a nice
uniform co-flow, so we have some other behavior which
could complicate this -- this -- what we see in our
test.
Okay. So what we did see, we did see
downcomer thermal stratification under the flowing
case. And we had a presentation on that which
describes what we saw when we were injecting these
plumes for the co-flow case and why that data looks
different, the downcomer profile looks different than
what we saw for --
CHAIRMAN WALLIS: Actually going db/dz
when you have these very irregular plumes must be a
little bit awkward. It's not as if you just have a
cone.
MR. REYES: Yeah. We --
CHAIRMAN WALLIS: You measure the db/dz,
this thing is swirling around.
MR. REYES: That's right.
CHAIRMAN WALLIS: So you had to -- do you
actually measure db/dz some --
MR. REYES: No, not with --
CHAIRMAN WALLIS: So this statement here
is inference from the theory.
MR. REYES: Inferred -- is inferred from
theory. It's inferred from theory.
CHAIRMAN WALLIS: Okay. I thought you
meant you had measured it.
MR. REYES: Yeah. That would -- that's my
dream.
If we could set up a -- this would be
simple enough to set up with a flow visualization. I
think it's something that --
CHAIRMAN WALLIS: With one plume?
MR. REYES: Just one plume, yeah. We
wouldn't see --
CHAIRMAN WALLIS: Once you start --
MR. REYES: Then grow to two plumes --
CHAIRMAN WALLIS: Once you start setting
up swirlies in the downcomer, all the plumes are going
to start wandering around.
MR. REYES: "Swirlies," I like that.
Okay. There's an appendix I have added to
this which deals with a little more details of what we
were seeing as far as the plume stratification or cold
leg -- excuse me -- cold-leg influ- --
CHAIRMAN WALLIS: Well, let's see. So far
it's kind of qualitative, isn't it? I mean I didn't
get anything I can grasp which is something I can use
yet.
MR. REYES: Right.
CHAIRMAN WALLIS: But you're going to give
us something which is more quantitative.
MR. REYES: Right. Now we're getting into
some more of the measured -- what we're seeing for the
different scenarios, the flow case and the stagnant
case.
So this is a stagnant loop case. We did
Tests Numbers 4, 5, and 6. We had the system
essentially stagnant. And we're injecting cold water.
There was no core heat. The system was hot initially.
We had all of our structure. Everything was hot in
that pressure. And we just started injecting cold
water into the cold leg and watch it spill into the
downcomer.
What this represents are all the
temperatures, and that's this part. Okay. These are
all the temperatures that lie underneath cold leg 1.
So directly beneath it and to the sides of it,
everything that's underneath cold leg number 1, all
the way down to the 8 d mark.
And on the -- of course, we're looking at
a decay situation. And so at first glance it looks
like everything is fairly tight. And this is kind of
what you'd expect to see. But at the very beginning
what happens is that your density difference, of
course, is the greatest. Okay. So you inject this
cold plume into the stagnant media.
And what happens, of course, is as time
goes on you're cooling off the whole system. And so
your delta rho is getting smaller and smaller as time
goes on. So you expect that difference to narrow. I
mean the plumes are, in essence, -- it's as if the
plumes are getting weaker, because the ambient is
getting colder.
So we see the largest temperature
differences initially, and this could persist for a
while, but some of the thermocouples we'll read. So
here we are at that 275 or so. We're looking at only
about a 25-degree-Fahrenheit difference from the
coldest point in the plume -- or the coldest location
in the downcomer, which is the 1.3 d for a lot of
these tests, and then down to about 4 d -- well, this
actually goes only 8 d.
So we're not seeing a very large -- except
maybe at the beginning, we see 50 degrees or so.
So when you first come in and the system's
hot, you really see a large temperature difference.
So you'd expect --
CHAIRMAN WALLIS: You've got so many
colors it's very hard to figure out anything.
MR. REYES: Oh, yeah. Yeah, this -- yeah,
that's a good point.
What I'd like to -- what I'd just want to
point out is the shape of this thing. Okay. How well
it's -- how tight it is, okay.
So what you see is kind of the --
throughout the test, these --
CHAIRMAN WALLIS: Which are the worst
spikes? Which ones are they?
MR. REYES: These bottom -- so what you
see are these bottom --
CHAIRMAN WALLIS: Which are the worst
spikes in terms of where they are, the location? Are
they at 2, or 3, or 4, or 8?
MR. REYES: They're all above the 4 d.
CHAIRMAN WALLIS: They're at 3 maybe?
MR. REYES: No, 2.
CHAIRMAN WALLIS: At 2?
MR. REYES: At 2 and 1.
CHAIRMAN WALLIS: 2 and 1?
MR. REYES: And that 1.3, it jumps and
sometimes it mixes back up. So you see -- you'll see
it come on and off basically. So we're still looking
for the --
MR. KRESS: And like David said, those
spikes don't matter. You can just ignore them.
MR. BESSETTE: I think, you know, to me
this one plot says that we have no plume problem.
MR. REYES: Yeah. It says it's relatively
tight all the way through. And you'd expect it to get
tighter --
CHAIRMAN WALLIS: Well, see if you had
bottom injection in that of ACC or -- you might have
a problem. The temperature differences would be much
bigger.
MR. REYES: Yeah. If the concern is
duration of temperature.
CHAIRMAN WALLIS: Yes.
MR. REYES: And when we look at the
flowing case --
CHAIRMAN WALLIS: And also magnitude. I
mean you're actual driv- -- your different temperature
differences in the cold leg aren't that big, anyway.
MR. REYES: That's right. Right. When we
look at the stratification.
So the rest of the plots, I'm going to
just skim through, but you see similar results.
CHAIRMAN WALLIS: Yeah. There are one or
two that look worse.
MR. REYES: Okay. That one was a little
bit more severe, but...
Again these are -- typically it's early on
in the transient, so we can just go through. So let's
jump over to the flowing case. So we'll jump over to
the natural circulation case.
So those tests represent essentially --
I'll say essentially stagnant plume. Again, what we
saw when we -- as soon as we injected it generated
initial loop flow which gradually stopped, so it was
hard to get -- for the tests that we did, because we
had system full, we saw a little bit of flow in the
cold legs initially, and that tapered off.
Okay. Now to the circulation case. This
was a series of parametric tests that we performed.
And what we had going on there is we used the core
power to generate a natural circulation flow rate
through the loops.
And then we varied our HPI in a stepped
fashion. And we produced 16 different cases, 8 cases
under cold leg 3 and 8 different cases for cold leg 4.
Okay. So they're on opposite ends of the plant under
different steam generators. So we were trying -- we
were watching both of these for the two situations.
And the idea is that we would hold it for
a small period of time and we would observe what type
of downcomer behavior, what type of plumes we would
generate.
So what you can see here is for one of the
more severe cases here about 200 kilowatts of core
power, which corresponds to about two-percent decay
heat for us and about half a gpm, which is about
50-percent HPI flow for that one leg.
We get about 300 -- about a 30-degree
temperature difference. And these top lines here,
again it's hard to see with this, but the bottom ones
are typically the 1.3 right at the outlet. And the
tops, it's kind of a merger of the -- between 4 and 8.
So we're seeing that the plumes have
essentially dissipated by the time you get to 4. And
I can show you individual plots with just -- which
would be a little bit easier to see, but this shows
all the data.
So that was -- the most severe case was
under 4, I think. Let's see -- no, so 3 was the more
severe case.
So we're seeing here only about 20 degrees
of temperature difference --
CHAIRMAN WALLIS: So 8, I can't see where
8 is because of where the color is.
MR. REYES: Oh, okay. Yeah, 8 merges all
the way up on top here. Yeah, 8's up on top.
And I can certainly -- we can pick any one
of these and blow it up.
CHAIRMAN WALLIS: And those are much less
than the stratification in the cold leg which you
showed us on another figure.
MR. REYES: Correct.
CHAIRMAN WALLIS: Much less.
MR. REYES: So the same test. First we
looked at the cold leg and then we looked at the
downcomer.
CHAIRMAN WALLIS: At most 30 percent or
so.
MR. REYES: Right. So we are seeing --
now the difference in this, in terms of a scenario --
okay. So if you have a reactor plant and you're
concerned about cooling at a -- for duration, well,
potentially in this situation you can have natural
circulation through a core. The core's producing hot
water. And so you're continually feeding hot water to
the cold leg.
So the difference between this and the
stagnant case is that in the stagnant case the plumes
will decay over time because the whole system is
cooling down. In this one you're using the hot water
from the cold leg to essentially keep your downcomer
warm. And so you can make these plumes persist a lot
longer. I think that's just an important point. But
we're seeing that temperature difference is not very
large.
Okay. And so if you get to even much
higher flow rates or -- with pumps on, it's not going
to be a big effect. But the difference is that you
can make these persist a lot longer because you've got
flow. You're replenishing your downcomers with hot
water. So that's the big difference there in terms of
an integral system.
CHAIRMAN WALLIS: Well, how big a
temperature difference do we need to see for Dave
Bessette to get worried?
MR. ROSENTHAL: Well, yeah, if I can
interject, it's not a question of Dave Bessette
getting worried but Nilesh Chokshi getting worried.
And we're talking -- what were you saying, 25 c, some
numbers like that.
CHAIRMAN WALLIS: So 25 c is bad?
MR. ROSENTHAL: Well, I mean --
MR. BESSETTE: Noticeable.
MR. ROSENTHAL: That's where it's starting
-- it will show up in their probabilistic fracture
mechanics case.
CHAIRMAN WALLIS: So these are less --
these are all less than 15 and maybe more.
MR. ROSENTHAL: Half.
CHAIRMAN WALLIS: In fact, the way you
worry about it is more like 2 or 3 c, isn't it?
MR. ROSENTHAL: Well, this is all good
news, but let me --
MR. REYES: Now --
MR. ROSENTHAL: -- let me back up a little
bit if I might just for a moment because I'm worried.
You know, if we go read this transcript two years from
now, the --
CHAIRMAN WALLIS: We should, yes.
MR. ROSENTHAL: -- we ought to put it into
a little bit of perspective.
We're talking about pressurized thermal
shocks. We're talking about low temperatures at some
higher pressure. We're talking about cases that would
have to persist for some period of time.
Roy Woods' probabilistic work will tell us
the probability of sequences. But here we're focused
on a small break LOCA sequence where the break is big
enough that you don't refill it and depressurize, but
not so big that you depressurize the system, so it
would be a case again. So it's sort of a
perverse-size break.
And then the issue comes up: Okay, we
know that under force-flow conditions we think that we
can predict things rather well. Under conditions
where you don't have force flow are you going to have
a problem with plumes and stagnant conditions and
whatnot.
CHAIRMAN WALLIS: Like you said, this was
an -- it's an artificial kind of break, but isn't it
--
MR. ROSENTHAL: No. It's just a purp- --
CHAIRMAN WALLIS: -- artificially made to
be a pretty bad break from the point of view of --
MR. ROSENTHAL: Of PTS.
CHAIRMAN WALLIS: -- of PTS.
MR. ROSENTHAL: Right. Right. But it's
--
CHAIRMAN WALLIS: And, of course, the risk
-- the risk-informed people will probably say it's
never going to happen anyway, so...
MR. ROSENTHAL: Well, he's going to give
us the probability of this thing some day. And then
when you throttle HPSI flow, then there's some
associated HRA numbers associated with that.
But from a thermal hydraulic standpoint,
it's still very interesting, so -- okay.
CHAIRMAN WALLIS: But you're trying to
look at the worst case, or something close to the
worst case in that?
MR. ROSENTHAL: Well, in this. I mean we
--
MR. REYES: Right.
MR. ROSENTHAL: Because I think that we
have confidence that in force-flow situations that the
codes will do a better job predicting --
CHAIRMAN WALLIS: So it is only if we
could say that this is something like the worst case
or very close to the worst case, and the temperature
differences here are so small that they don't
challenge in any way the vessel, then --
MR. ROSENTHAL: Okay. But stagnant --
CHAIRMAN WALLIS: -- we don't need any
PRA. We can forget it.
MR. ROSENTHAL: Well, fine. Okay. But --
CHAIRMAN WALLIS: And then we're really
happy then.
MR. ROSENTHAL: But the -- yeah. But the
stagnant thing was something of true concern, --
CHAIRMAN WALLIS: Yes.
MR. ROSENTHAL: -- so I'm glad that we --
CHAIRMAN WALLIS: Yes. Oh, yes.
MR. ROSENTHAL: -- had the experiments.
CHAIRMAN WALLIS: Yes, indeed. Indeed.
MR. ROSENTHAL: And --
CHAIRMAN WALLIS: Otherwise, you could
argue about it forever.
MR. BESSETTE: Yeah. And it's worth
noting, we didn't do these experiments by accident.
(Laughter.)
CHAIRMAN WALLIS: Well, I hope you don't
do many experiments by accident.
MR. ROSENTHAL: David keeps whispering
little insights in my ear about the timing, et cetera.
I mean --
CHAIRMAN WALLIS: Someday --
MR. SHACK: Again this is all for
Palisades. You know, you're going to have to somehow
do the calculations to convince us that it's this good
for everybody.
MR. ROSENTHAL: Okay. But this is where
-- and we'll do lots of calcs, but I mean this is
really very good benchmark stuff. And even if you
take the broader perspective, we have relatively
little data on steam line break experiments. So the
fact that we're generating these have a broader
applications in terms of --
MR. SHACK: Well, that, I guess, was
another question --
MR. ROSENTHAL: -- code validation.
MR. SHACK: -- is I assume you're getting
all this in a nice optical disk somewhere so someday
in the future if somebody wants to benchmark
calculations, this won't disappear.
MR. BESSETTE: It's all going on our
databank, yes.
MR. ROSENTHAL: Oh, so you can get it
right off the web.
MR. SHACK: Yeah, I could. From the NRC
website. Good luck. Or you're going to put it in
Adams.
(Laughter.)
MR. REYES: There is one qualifier that --
I just want to remind you, that this is -- we are a
reduced-pressure facility. And so as a result our
delta rhos over rho are somewhat less. What you'd
expect to see in Palisades is something on the order
of 10 percent or more.
CHAIRMAN WALLIS: That's when we get into
the question of when you get a big enough delta rho do
things change significantly, or is it still the same?
MR. REYES: Is there a transition
somewhere. So that's just a -- just as a qualifier.
We're fairly close. I mean .18 is
relatively typical. But I think for Palisades plant
we're going to see they start up at a higher
temperature, 570 and -- or hot leg 570, the cold leg
530. So they're 530, we're at 420. So there is some
difference.
CHAIRMAN WALLIS: But it's not as if
you're way off.
MR. REYES: Right. That's right. Okay.
CHAIRMAN WALLIS: And if you have a good
theory, then maybe extrapolating it that far isn't too
bad.
MR. REYES: That's not so bad, no.
Okay. So concluding with this one, what
we saw was that the downcomer plumes were -- we saw it
both for the stagnant case and natural circulation
flow conditions.
For the range of natural circulation flows
we examined from one and a half percent to four
percent decay power in this test number 3 and about 30
to 100 percent HPSI flow.
The plumes were, for all those cases, they
were well mixed by about four cold-leg diameters, is
what we saw. Okay. Everything was relatively back to
the ambient temperature.
CHAIRMAN WALLIS: How does that compare
with what we started out with? You know, we started
out with a theoretical plot for the infinite plume.
You said Rodi or someone has all these different
conditions.
Does yours fit in in any way with the
classical experiments with isolated plumes?
MR. REYES: With the -- we haven't --
right. We haven't done that comparison yet.
CHAIRMAN WALLIS: But this seems to be a
very rapid mixing.
MR. REYES: It's a very rapid mixing. We
haven't -- my first thought was to use the co-flow
work and the isotonic models to compare, just to see
if that makes a difference. But I haven't done that
yet.
But we have done calculations with REMIX
and with STAR-CD, our CFD codes, to try and see if we
can predict some of the meandering behavior in some of
the --
CHAIRMAN WALLIS: This is funny, because
jets are sort of well mixed by 10 diameters, aren't
they? And you'd expect plumes to go further because
the buoyancy is driving them. So it's kind of
surprising that four diameters are enough to --
MR. REYES: To mix this --
CHAIRMAN WALLIS: -- to wear out these
plumes.
MR. REYES: These are -- again, this is
for the flow rates that we're looking at. And, again,
for the CE Plant that they're a very low injection
flow compared to what we see in other plants, so...
Dave suggested that --
CHAIRMAN WALLIS: Well, it may well be --
excuse me -- but the actual plume, when it starts out,
is much smaller than --
MR. REYES: I know, just --
CHAIRMAN WALLIS: -- the whole cold-leg
diameter.
MR. REYES: Right. That's right.
CHAIRMAN WALLIS: For instance, this
spills out --
MR. REYES: That's right.
CHAIRMAN WALLIS: -- and maybe you should
take -- if you took a quarter of a cold leg diameter,
this would look like 16 diameters.
MR. BESSETTE: That's right, yeah.
MR. REYES: That's right.
CHAIRMAN WALLIS: Maybe that makes more
sense.
MR. BESSETTE: It's not evident that the
cold-leg diameter has any particular significance.
MR. REYES: Okay. Today --
CHAIRMAN WALLIS: Well, it's a way of
scaling things, isn't it? It's just...
MR. REYES: You're right. It's a good
idea.
Today when we run the test I'll need
someone take a look in the tank and I think, you know,
we'll be able to see as it pours into the sides of the
-- I think you're right. It doesn't fill the pipe.
CHAIRMAN WALLIS: We'll see a plume?
MR. REYES: You'll see a plume. Well, one
person will.
CHAIRMAN WALLIS: Only one person, no
independent check on that?
MR. REYES: That's right.
CHAIRMAN WALLIS: One person goes in the
tank and looks through the wall?
MR. REYES: One in the tank.
Okay.
CHAIRMAN WALLIS: Do you need a volunteer?
MR. REYES: Well, we've got two.
Okay. Are any other questions on what
I've presented?
(No audible response.)
MR. REYES: If not, then we'll move to the
next presentation.
Brandon is going to talk, Brandon Haugh
will be talking a little bit about what we saw for
these force flows or natural circulation flows, what
we saw happening in the downcomer that was a bit
different than when you had the stagnant flow.
MR. HAUGH: Just give me one second here.
DOWNCOMER THERMAL STRATIFICATION IN APEX-CE
MR. HAUGH: Good morning, everybody. My
name is Brandon Haugh. I'm a graduate student in the
Department of Nuclear Engineering. I'll be giving you
a presentation on downcomer thermal stratification we
observed in our CE Tests in the APEX facility.
I am going to talk about a description of
what downcomer thermal stratification is. I have a
diagram and some tests from the IVO facility in
Finland, observations of what we saw in our test
facility. And we're also going to run another one of
those transient temperature maps that we saw in a
previous presentation, and it will help to easily
demonstrate what thermal stratification is. And then
I'll come to a few conclusions.
The figure here is rather dramatic. It
doesn't actually look that stratified. This is just
for an appearance of -- it looks good in black and
white, so we'll leave it at that.
And the plume isn't this concentrated. It
obviously spreads and dissipates some. But you'll see
that there's co-flow of velocity in the downcomer and
velocity in the plume when we see a stratified layer
in the lower portion of the downcomer.
We observed in our tests that this
occurred in the presence of natural circulation flow.
It didn't happen in any of the stagnant cases. And it
seems the co-flow of the downcomer fluid stream in the
plume reduces the mixing and seems to permit the onset
of downcomer thermal stratification.
It seems to help confine the plume, and it
seems to just go to the bottom and start cooling the
bottom and working its way up.
These figures here are taken from the IVO
facility. I would say the full name, but I would
probably butcher it.
This is from Test Number 102. In some of
the tests they used photography so it was just not all
artist rendition. But for this test they had one
injection in one cold leg and then cold-leg flow in a
different cold leg. But it was a rather high flow
rate, so you can see it's pretty dramatic, the
stratification they see --
CHAIRMAN WALLIS: Well, that plume isn't
mixing much at all, is it, unless I'm mistaken.
MR. HAUGH: Yeah, exactly. There's a 66
gallons per minute cold-leg flow and 6.6 gallons per
minute HPI flow. So this is much larger than we see
in our facility.
CHAIRMAN WALLIS: What we see on the left
is --
MR. HAUGH: Yeah. You'll see the plume
almost penetrates fully to the bottom.
CHAIRMAN WALLIS: The plume doesn't seem
to spread at all.
MR. HAUGH: Yeah, exactly. And that's
kind of the argument of the co-flow case. It seems to
help confine the plume.
And you'll see the stratification. I mean
it's rather dramatic because of the dye. You can't
really tell kind of how much it mixed at the bottom.
CHAIRMAN WALLIS: Then you say,
"Temperature gradient observed in the downcomer" are
similar?
MR. HAUGH: Yeah, we see similar in our
facility, but it's not quite as dramatic.
CHAIRMAN WALLIS: What's this? I would
think that you'd actually get cold temperatures lower
down with that, a plume like that.
MR. HAUGH: Yeah, that's what we see. You
know, in the transient temperature map you'll see that
we don't seem to catch the plumes due to maybe our
thermocouple spacing or it jumping and may be sticking
to the core barrel side. But we definitely see the
stratification. Okay.
For the test that we ran, an integral
systems test, and some of the separate effects test,
this is just kind of a map of what we did and what we
saw related to downcomer thermal stratification.
The first three, 4, 5, and 6, were the
stagnant cases with no cold-leg flow. And we observed
basically no downcomer thermal stratification.
In tests 7, 8, and 9, those were the small
break LOCAs, where we had natural circulation flow in
at least some of the cold legs, we did observe some
downcomer thermal stratification.
CHAIRMAN WALLIS: And by "thermal
stratification," you mean temperature as a function of
z, or as a function of Froude? So the pool --
MR. HAUGH: Yeah, in an axial measurement.
So from the bottom of the lowest point --
CHAIRMAN WALLIS: -- a pool of cold s- --
MR. HAUGH: -- in the downcomer to the
top.
CHAIRMAN WALLIS: Why doesn't that wash
out when you have circulation?
MR. HAUGH: That's -- we don't quite -- I
don't quite understand that. I think Dr. Reyes might
be able to field that.
MR. REYES: What we've got is a relatively
low natural circulation flow. And that's introducing
the hot water into the top of the downcomer. And that
water is -- at the bottom you do see some of that,
that mixing occurring. And so it's constantly
replenishing that mixing region which is a little bit
lower in the downcomer.
CHAIRMAN WALLIS: So the hot water is
getting out by mixing with the cold, presumably.
Otherwise --
MR. REYES: Right. That's right.
CHAIRMAN WALLIS: -- it would --
MR. HAUGH: You'll notice that the
stratification we observed isn't very significant in
terms of the delta T, from 8 diameters to 1.3
diameters; in tests 7, 8, and 9 it was between 35 and
40 degrees Fahrenheit.
In Test Number 10 -- that was a
combination test of a safety relief valve in the
pressurizer and an atmospheric dump valve on the steam
generator side -- we saw a little, of a slightly less
stratification, but it was observed.
In the hot zero power main steam line
break due to the -- we have no core power basically.
Well, it was very low, like 45 kilowatts, we weren't
replenishing the hot water in the top of the
downcomer. So we didn't observe the thermal
stratification. It appeared to be relatively well
mixed.
In Test Number 12 in the main steam line
break from full power while the steam generator was
blowing down, the downcomer was relatively well mixed
because the cold legs feeding in were relatively cold
from the broken side.
But after the steam generator finished
flowing down and we started reheating the plant after
we repressurized, we saw the onset of thermal
stratification because we started feeding hot water
into the top of the downcomer. But it was very minor
and not significant delta T from the top to the
bottom.
This is a snapshot of one of the transient
temperature maps from Test Number 9, which was a
stuck-open safety relief valve on the pressurizer.
This is the initial conditions. We can see it's
relatively all the same temperature. I'll move
quickly to the next slide.
This is 3500 seconds into the test, and
it's pretty obvious to see. You can see the
stratification in the downcomer. It's easy to note
that it's, you know, only 30 degrees Fahrenheit, but
it is there.
Later on in the test we can see that the
bottom is cooling up, but we're still replenishing hot
water due to the core power to the top, so it's
stratified. But the stratification layer is getting
much closer to the cold legs at this point.
CHAIRMAN WALLIS: With the stratification
that high it really kills the plume, doesn't it? The
plume goes right into the --
MR. HAUGH: Yeah, we --
CHAIRMAN WALLIS: -- stratified thing, and
then --
MR. HAUGH: In the facility and in some of
the plots that we don't have. But when we play the
transient temperature map, you can -- you won't even
notice that there's a plume there. It just seems that
you just see the stratification, and that's about all.
And then for Test Number 6, which was one
of the stagnant cases where we didn't observe thermal
stratification, this is about 800 seconds into the
test, you can see that it's all relatively well mixed.
And at this point the delta rho over rho
is a -- well, the delta rho between the downcomer
fluid and the plume is relatively low, so you don't
really see much plume activity, either.
And now we'll play the transient
temperature map from Test Number 9. And I'm starting
about 2,000 seconds into the test, because that's when
it's easiest to see this onset of the stratification.
And you can see that the bottom is slightly colder
than the top. And it'll build as it runs.
You can also notice the stratification in
the cold legs from the injection. And it seems to be
pretty much -- above the Weir wall height it's
relatively constant temperature fluid. But due to the
side injection below the Weir wall height it's
stratified.
And when you've seen enough, just let me
know.
CHAIRMAN WALLIS: If you gave each one of
those dots a note you could play music.
(Laughter.)
MR. HAUGH: Next time.
And as you can see, it's starting to cool
as we go up. That's probably... I mean would you
like to see more? It's entertaining.
That's probably enough, Kent.
Just click on "Resume Slide Show" up
above. There you go. Okay.
Now that we've seen that, I hope it was
entertaining.
So the conclusions about downcomer thermal
stratification, DTS, we observed it in the APEX
facility in tests where we had natural circulation.
And we noticed it did not occur in the stagnant loop
cases or when we had a very high cool-down rate, such
as in a main steam line break.
But after the blowdown ceased in Test
Number 12, when we started reheating, we saw the onset
of thermal stratification.
We have come to the conclusion that a
probable mechanism for downcomer thermal
stratification is the co-flow of the downcomer fluid
stream and the plume, which tends to preserve the
plume and helps it get to the bottom.
And also we have the replenishment of the
hot fluid from the core power heating up the top of
the downcomer.
CHAIRMAN WALLIS: There's more to it than
at that time, though, isn't there? I mean it somehow
has to go through almost a shock wave. The plume
comes down and then it all mixes up in this colder
thing at the bottom.
MR. HAUGH: Yeah, there's a mixing -- with
a thermal stratify decay there's like a mixing --
there's probably some penetration.
CHAIRMAN WALLIS: I would think you'd have
to explain that. It's not just a fact that the plume
is helped by the co-flow. Why does it stop? Why does
this mixing occur at that level?
MR. HAUGH: That's a good question.
This was just a preliminary conclusion.
There's obviously more mechanisms present that --
there's the plume interaction with the thermal
stratified layer also hitting the core barrel.
There's several other things that will probably need
to be examined to further define this phenomena.
That concludes my presentation. Is there
any more questions I can field?
(No audible response.)
MR. HAUGH: Okay.
CHAIRMAN WALLIS: Well, we're looking
forward to a really good theoretical model.
MR. HAUGH: And the work is in progress.
(Aside comments off the record.)
CHAIRMAN WALLIS: I think we have one more
presentation before the break.
MR. BOEHNERT: That's correct.
REMIX CALCULATIONS OF APEX-CE TESTS
MR. YOUNG: My name is Eric Young. I'm a
graduate student here, a graduate student of Dr.
Reyes. I'd like to take this opportunity to thank Dr.
Reyes for giving me this opportunity for presenting in
front of the council -- or the Committee.
I'll be presenting on the REMIX
calculations or predictions on the STAR- -- or on the
APEX-CE facility. The presentation will progress in
the following manner.
I'd like to describe the objectives of the
study; go through a description of the REMIX model; a
description of the APEX-CE stagnation tests that we
did a comparison with, these being Tests 4, 5, and 6;
some insight into the effects of the core barrel heat
transfer and effective thickness; and the recommended
summary -- or summary and conclusions.
The objectives of the study was to
benchmark STAR-CD against the integral test facility,
this being the APEX-CE facility here at Oregon State
University; identify any of the code limitations in
predicting downcomer and well-mixed temperatures at an
integral test facility; --
CHAIRMAN WALLIS: You said STAR-CD, did
you, or REMIX?
MR. YOUNG: Did I?
MR. REYES: Yeah, REMIX.
MR. YOUNG: REMIX. Sorry. I worked on
both things. I might put the two together sometimes.
...benchmark REMIX against the APEX-CE
facility; and assess the applicability of the code for
integral system geometries.
The REMIX Computer Code is used for
calculating well-mixed core inlet temperatures along
with downcomer temperatures in any specified locations
below the cold-leg injection into the downcomer.
It's based on the regional mixing model
originally designed by Dr. Theophanus and is described
in the following figure.
CHAIRMAN WALLIS: Dr. who?
MR. ROSENTHAL: Theophanus.
CHAIRMAN WALLIS: Theophanus?
MR. YOUNG: And it's described in the
following figure. This is out of the REMIX Manual.
It assumes that the cold stream originates
at the high-pressure safety injection site, forming a
cold stream along the bottom of the cold leg, which
then flows towards the downcomer and the loop seal
portions of the primary system.
At these two jump locations it generates
a buoyant plume, which decay into the loop seal and
into the downcomer. Most of the mixing -- or the
mixing is most intense at certain mixing regions, and
these are the regions that REMIX calculates mix and
entrainment, which it uses to then determine the cold
stream temperature that enters into the downcomer, and
it also does a global system calculation for the
well-mixed temperature inlet to the core.
CHAIRMAN WALLIS: Is there some kind of
coefficients which describe the mixing --
MR. YOUNG: Yes. Yes, sir. This is --
CHAIRMAN WALLIS: -- which are different
in the different regions and determined in some
empirical way?
MR. YOUNG: I'm sorry?
CHAIRMAN WALLIS: Are they determined
empirically, these mixing coefficients?
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: So the data are made to
fit these experiments, or...?
MR. YOUNG: It was pitted against --
CHAIRMAN WALLIS: Some similar...?
MR. YOUNG: It was originally validated
against the Creare one-fifth, I think, scale test, and
it was again modified using some mixing experiments at
Perdue University, I believe.
MR. SCHROCK: Could you go back and
explain this pump? What is this description of the
pump all about? I don't understand it.
MR. YOUNG: Okay. In REMIX you're able to
specify a certain pump volume in the pump heat
transfer area to take into account any contribution
that the pump material or structure would have in
contributing to the temperature of the system, of the
coolant inside the system.
It's an option allowed in REMIX.
MR. SCHROCK: I don't understand what this
diagram is intending to convey with a stratification
situation. It looks like it's spilling over in a
reverse flow and then an arrow showing something else
moving forward.
MR. YOUNG: Okay. In the calculation of
the amount of mixing entrainment from the cold stream,
there needs to be a countercurrent hot stream flow
from the downcomer and the loop seal portion.
The countercurrent flow from the loop seal
portion is generated from the inventory that's in the
loop seal during the injection and is then mixed with
a cold stream.
The cold stream was injected from the
high-pressure safety injection location and flows
towards both ends of the cold leg.
The hot stream that finds the mix and
entrainment region flows from the upper downcomer --
the upper plenum in the downcomer into the top of the
cold leg. And that finds the mix and entrainment at
the falling plume location, which is the high-pressure
safety injection location.
CHAIRMAN WALLIS: So these MR 1, 3, and 4
are the areas which have some kind of a coefficient to
describe mixing, which is different for each region?
MR. YOUNG: Yes, sir. These are regions
that are assumed to be determinant in how much mixing
occurs between the cold stream and hot stream. Any
mixing at other locations of the system are
negligible.
It's assumed that there is a stationary
interface between the cold and hot streams on -- in
any locations, other than these mixing regions.
MR. SCHROCK: MR5 is insignificant or
significant?
MR. YOUNG: MR5 is calculated in REMIX,
and the only insignificant mixing region is MR2. It
isn't shown in this map, but it's between the high-
pressure safety injection and the vessel in that
length of cold leg.
Okay. Some of the limitations of the --
MR. SCHROCK: The code chops this up into
elements and calculates something? It's kind of a
sketchy description of what the code is.
CHAIRMAN WALLIS: I presume it has nodes
and --
MR. BESSETTE: Volumes.
CHAIRMAN WALLIS: -- cold volumes.
MR. YOUNG: Yeah. In the -- in order for
REMIX to calculate the temperature transients at the
downcomer locations and the well-mixed temperature
transient at the core inlet, it is required that you
specify certain volume, the total volume,
participating volume, of the system; the total mixing
volume of the system, which is considered to be in
well-mixed conditions.
Any material structures that are masses
indirectly specified by their wall thickness and their
material properties, which is the conductivity and the
diffusivity of the material structures, which allows
any conduction calculation to be calculated along with
the transient temperature of that material.
MR. SCHROCK: Okay. I'm exhausted.
MR. YOUNG: Is that sufficient?
CHAIRMAN WALLIS: I presume there's a
document somewhere that describes REMIX.
MR. YOUNG: Yes, there is, sir. I have
two REMIX Manuals, if you'd like to look at them.
MR. BOEHNERT: We can get that for you as
soon as you get back to the office.
CHAIRMAN WALLIS: Yet another code to
review.
(Laughter.)
MR. SCHROCK: I'm not sure I'd like to
review it, but I'd like to be reminded of something
that I did look at more than 10 years ago. I'm not
getting very much from this.
MR. BOEHNERT: I'll get you a copy.
MR. BESSETTE: You might say that each of
those mixing regions that are labeled, you might say,
are distinct calculations that REMIX does.
MR. YOUNG: It predicts -- a plume is
assumed to be developed within two diameters down from
the cold-leg injection. This temperature is
calculated and many other specified locations in the
downcomer which are desired are then calculated from
this location.
And then a lumped parameter calculation is
made to determine the well-mixed core inlet
temperature, which is simply an exponential decay to
a stable temperature.
Some of the limitations of the code in the
version that we're using. The version that we used
for the comparison is the code -- is the 1986 version.
And some of the limitations in this code is that it
was not designed to predict the effects of multiple
plume interactions.
It's not designed to predict temperatures
or calculate temperatures in the presence of cold-leg
flow. The only output that's available from this
REMIX is the cold centerline temperatures and the core
inlet well-mixed temperature.
You can't specify locations that are any
different as a myth of location, so -- in between two
of the cold-leg injections; it's only along the
centerline.
And it doesn't predict plume velocities or
any other flow characteristics. It uses a fixed-heat
transfer coefficient, which we'll see later to be
negligible or really not that important in the
calculation.
And it does a 1-D conduction heat-transfer
analysis in the metal structures.
The three tests that we compared with
REMIX were tests OSU-CE-004, 5, and 6. The
superficial Froude number for these tests range
anywhere from .019 to .0402.
The temperature difference between the
high-pressure safety injection and the primary
inventory gave delta rho over rho values of .18.
Between the three tests, the first test,
004, only had one injection location, where Tests 5
and 6 used all four high-pressure safety injection
systems.
CHAIRMAN WALLIS: Qhpsi is per injection?
MR. YOUNG: I'm sorry, sir?
CHAIRMAN WALLIS: Qhpsi is the flow rate
per site, so we have --
MR. YOUNG: It's per site.
CHAIRMAN WALLIS: Okay.
MR. YOUNG: And in the case of four
high-pressure safety nozzles being operated it's an
average flow rate between the four, because there is
some variance between them.
The input data to describe the APEX-CE
facility model in REMIX was created by specifying the
volumes for the -- in the one high-pressure safety
injection case the entire downcomer and lower plenum
was considered as a portion of the total volume.
In the four high-pressure safety injection
case, the lower plenum and downcomer region was
partitioned into four equal volumes.
The material properties for the
structures, which I mentioned before were the
conductivity and diffusivity, were specified: The
HPSI flow rates and relative flow temperatures or
fluid temperatures; initial system temperature; heat
transfer areas in each of the regions along with their
respectful heat transfer coefficients.
CHAIRMAN WALLIS: You input those; they're
not calculated in some way? You just put some number
in?
MR. YOUNG: These are variables that you
have to put in to describe the facility to REMIX. So
you have to specify a certain amount of volume that
you consider to participate in this calculation. And
a well-mixed volume, which I mentioned, is considered
to be well mixed during --
CHAIRMAN WALLIS: Well, how do you figure
heat-transfer coefficient?
MR. YOUNG: The heat-transfer coefficients
that I calculated were calculated using analytic plume
velocity in the downcomer region. And in the cold leg
I used the flow rate over the area of the cold stream
for that velocity.
CHAIRMAN WALLIS: These are just at the
wall; they're not between streams, aren't they? Just
at the wall?
MR. YOUNG: This is not between the
streams.
MR. SCHROCK: I thought David told us
that's essentially infinite heat-transfer coefficient.
MR. YOUNG: That is -- we did a heat-
transfer coefficient sensitivity, which I'll show you
towards the end of this presentation, and it's fairly
insensitive to any changes in heat transfer.
In calculating -- or in properly
describing the REMIX facility or the APEX-CE facility
with REMIX, we considered the core region to be
participating in the heating of the downcomer fluid.
As you can see in the figure in front of
you that there is a ceramic annulus, which is of
nonuniform thickness, around the core. It's directly
in contact with the core barrel and supplies energy to
the core barrel for heat transfer to the downcomer.
In order to maintain a correct description
of the heat conductance in our -- during our test, we
used an equivalent material which would include the
downcomer stainless steel and the ceramic of the
reflector.
We considered these to be homogeneously
mixed. And we calculated an effective thermal
conductivity for this material to specify in REMIX.
And then we increased the core barrel thickness to
include the thickness of both of the materials.
This is to account for any energy stored
in the reflector that is available for heat transfer
to the downcomer fluid.
Now I'd like to show you some of the
comparisons between these four -- three tests.
CHAIRMAN WALLIS: So even in the transient
you can use just conduction here to the core barrel
reflector?
MR. YOUNG: Yes, sir. We gave -- we
specified an effective diffusivity so that REMIX would
be able to calculate the temperature transient of the
material and from that the heat transfer to the
downcomer fluid.
And the heat-transfer coefficient was
again the calculated heat-transfer coefficients used
in the plume velocities.
In the first graph, this is the core inlet
well-mixed temperature for Test 4. This is with one
HPSI being operated.
As you can see, REMIX underpredicts the
temperatures at all the locations within 40 degrees of
the actual well-mixed core inlet temperature.
CHAIRMAN WALLIS: Well, the well-mixed
temperature is simply mixing the fluid and having some
heat transfer from the wall to change it?
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: And so the only variable
really is the heat transfer from the wall. The mixing
is this first law conservation of energy.
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: So what, so the
difference is presumably in getting the heat-transfer
coefficient right.
MR. YOUNG: It is in correctly modeling
the amount of heat transfer that the --
CHAIRMAN WALLIS: Right.
MR. YOUNG: -- core region contributes to
the downcomer.
In REMIX calculations and in the Manual
it's recommended that a core barrel thickness be
specified. This is specified in any previous
literature with validation. Only the thickness of the
core barrel itself was described.
It's seen in some of these plots that a
portion of the core region and energy stored in the
material and possibly even some of the primary
inventory in the core region is contributing to heat
transfer through the downcomer region.
This next plot is a temperature comparison
between 1.3 cold-leg diameters below the cold leg
injection into the downcomer for the case of one
high-pressure safety injection being operated.
Again, REMIX initially underpredicts the
temperature of the plume at the location and
throughout the entire test. It's within 60 degrees of
the actual calculations.
CHAIRMAN WALLIS: But REMIX is predicting
far bigger temperature differences than you actually
get.
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: So it seems to be way
off, rather.
MR. YOUNG: Well, it's much more accurate
in predicting a four high-pressure safety injection
case than it is a single high-pressure safety
injection case.
I think this is due to -- maybe I'm not
completely understanding the mixing volumes that are
participating in the one case.
This is for four cold-leg diameters below
the cold-leg injection into the downcomer. And again
REMIX underpredicts the temperatures within 40
degrees.
The same calculations were carried out
previous to increasing the core barrel thickness to
include the mass.
CHAIRMAN WALLIS: Excuse me.
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: I think that Professor
Reyes showed us that this 4-D temperature is very
close to the mixed temperature.
MR. YOUNG: In the facility?
CHAIRMAN WALLIS: Yeah, right. So that,
in fact, REMIX is getting the mixed temperature wrong.
It's really giving it a tremendous difference. It
very well matters really. So the difference between
the 4-D temperature and the mixed temperature, is what
you're worried about, which we know to be very small.
So REMIX --
MR. YOUNG: After approximate- -- yes,
sir.
CHAIRMAN WALLIS: -- on that basis is way
off, isn't it?
MR. YOUNG: Yes, sir.
MR. BESSETTE: And you'll notice the
offset seems to occur from time, from your initial
time.
MR. YOUNG: We noticed in the stagnation
cases that after approximately 500 seconds into the
test the plume had seemed to have been completely
diminished.
So at that point REMIX doesn't calculate
any other flow configurations which would enable it to
determine whether or not the plume existed.
The next --
MR. ROSENTHAL: Yeah, if you'd just flip
back one slide. So this is the centerline, roughly?
This is along the centerline?
CHAIRMAN WALLIS: It probably gets too
cold, the plume just coming out of the pipe.
MR. BESSETTE: That's right. I think
that's what the problem is. If you -- I -- you know,
what REMIX does for this, if you go back to Viewgraph
Number 4, what REMIX does for mixing Region 3 is
arbitrary. And I think it's getting that --
CHAIRMAN WALLIS: It doesn't mix enough.
MR. BESSETTE: Yeah.
MR. ROSENTHAL: Okay. But, you know, last
week and the week before we're running RELAP and REMIX
calculations. And if you look at 4,000 seconds you
see about a 50F difference on that graph. And so
that's now something that we would start to see in the
fracture mechanics stuff.
So the offset is important to us. And we
had, you know, RELAP Code and REMIX Code. And we're
scratching our heads whether we should believe any of
it.
So now we've got some REMIX versus some
experimental data, which will allow us to come to some
conclusions about what we should do with the REMIX.
CHAIRMAN WALLIS: So your strategy might
be --
MR. ROSENTHAL: So I'm just pointing out
that, you know, where this fits in the grander scheme.
CHAIRMAN WALLIS: So your strategy might
be to try to fix up REMIX to represent these
experiments.
MR. ROSENTHAL: Or shuck it and go to
CFD's stuff --
CHAIRMAN WALLIS: Or go to something else,
yes.
MR. ROSENTHAL: -- that you couldn't have
done in the mid-'80s.
CHAIRMAN WALLIS: All right.
MR. YOUNG: If you're able to --
CHAIRMAN WALLIS: If no one else wants to
speak, is it your objective to just run REMIX and see
how it does, or is it to fix up REMIX to be more
realistic?
MR. YOUNG: Well, it was originally to use
the recommended REMIX or the -- what REMIX recommended
for the volumes and structural materials to see how
accurately it could determine integral test facility
with either a heated core region --
CHAIRMAN WALLIS: So then an assessment of
REMIX, yeah.
MR. YOUNG: -- or mix.
An assessment. And in doing this,
describe any limitations and fix these limitations
that REMIX might have. And this was -- in my first
attempt -- was to increase the core-barrel thickness
to include any of the mass.
CHAIRMAN WALLIS: Right, right. So you
are fixing it up, as well?
MR. YOUNG: I'm trying to -- I'm trying to
find if REMIX is applicable to an integral test
facility, since there are only basically two different
conditions that can change the fluid temperatures.
And that's the mixing and the thermal conduction from
the wall.
CHAIRMAN WALLIS: Right.
MR. REYES: Now there appears to be
limitations to what we're seeing in the experiment
compared to what REMIX can predict.
You know plume interactions and things
like that REMIX cannot do, so Eric's been working, of
course, with STAR-CD. And you'll hear another talk on
that, trying to see if we can come up with some better
-- better tools to predict the behavior, because we
think there are some limitations to what the code was
designed to do.
It was really for a stagnant condition, a
single plume. And he'll show you some more plots
looking at one plume versus four plumes. But we see
big differences in the behaviors -- in the calculation
versus the behavior.
MR. YOUNG: As we've seen in a lot of the
CFD calculations or we will see later on today, the
ability for REMIX to predict the location of the
coldest transient is almost impossible with REMIX,
because it predicts only centerline temperatures.
We're finding our coldest downcomer or our
coldest vessel-wall temperatures to be located between
two interactive plumes after they have merged. So
it's not really applicable to the case where there's
more than one injection.
CHAIRMAN WALLIS: Now you're going to show
us all the graphs. They all look very similar.
MR. YOUNG: I'll just show you a couple
more, and I'll go on to the --
CHAIRMAN WALLIS: Right. Then go on to --
MR. YOUNG: -- transfer system.
CHAIRMAN WALLIS: -- improvements, such as
bringing in the core barrel, or whatever.
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: Thank you.
MR. YOUNG: I'll just flip through these
quickly then to show you that REMIX does underpredict
because it is, indeed, calculating that there is a
plume there. That's inherent in the code.
Pick one of the well-mixed temperatures
here. Here's a case where four injection -- HPSI
injections were being operated. You can see that it
predicts the well-mixed temperature or the coolant
temperature much more accurately.
CHAIRMAN WALLIS: If it starts right it
seems to do better.
MR. YOUNG: Yes.
CHAIRMAN WALLIS: It has to start right at
the top.
MR. YOUNG: Yes.
Yet it still underpredicts the downcomer
locations. These are more accurate. They're still
under by 25 to 30 degrees.
One of the things that we found very
important in these REMIX calculations was the effects
of the core barrel or the core region or any materials
in the core region that had stored energy that could
be supplied to the downcomer.
We did a heat-transfer coefficient
sensitivity study along with downcomer thickness,
wall-thickness sensitivity.
In this first slide we varied the heat-
transfer coefficient from 100 watts from u squared
degree Kelvin up to 6,000 watts from u squared degrees
Kelvin.
You notice that the difference in
temperatures calculated for the core inlet temperature
are very negligible. They're only within a couple
degrees, maybe 5, 10 degrees.
This, indeed, is the case because, as
David Bessette had mentioned, that we are
conduction-limited and there are the skin effects of
the heat transfer, removing the energy from the skin
of the materials.
Yet when we vary the thickness of the core
barrel to include the mass of the reflector, we see
that we calculate temperature differences up to 35 to
40 degrees Fahrenheit. So the material or the energy
stored in the core region is, indeed, important in
these calculations.
MR. KRESS: So did you vary the mixing
rate in the downcomer?
MR. YOUNG: That's not available in the
code that I'm aware of, sir. I think it's hard-coded
in it. They have I think a linear segregation of the
plume from the injection location down to two
diameters.
MR. KRESS: That would -- that would
certainly be a way to bring the codes -- the cores
together.
MR. YOUNG: If we were able --
MR. KRESS: Yeah.
MR. YOUNG: -- to understand a little bit
more of how the mixing occurred and to introduce that
into the code.
CHAIRMAN WALLIS: Now there's Mixing
Regions 1 and 3 you need to do right. You have the
starting condition right.
MR. KRESS: Well, he also gets -- needs to
do it in the plume, because --
CHAIRMAN WALLIS: Yeah.
MR. KRESS: -- because that's why it cools
-- cools off faster than --
CHAIRMAN WALLIS: But by now he's probably
got four tuning coefficients. He should be able to do
very well.
MR. KRESS: Yeah. You ought to be able to
match it exactly.
(Laughter.)
MR. KRESS: He's only got one for the
plume.
MR. YOUNG: So, in summary, the REMIX
model was developed and applied for three of the
stagnant loop conditions. In the comparisons for the
one high-pressure safety injection operation, REMIX
underpredicted both the core inlet and the downcomer
locations.
For the cases of tests 5 and 6 where four
high-pressure safety injections were operated, the
predicted -- or calculated temperatures were much more
accurate, yet REMIX still underpredicted all of them.
One of the reasonings behind the
underprediction of the 1.3 location is again that we
believe the cold stream entering into the downcomer
jumps over this location, so the temperature or
thermocouple in the APEX facility isn't reading the
cold stream temperature. It's rather reading the wall
-- near-the-wall temperature. And --
CHAIRMAN WALLIS: So you're saying in
reality the plume jumps and REMIX doesn't consider
this?
MR. YOUNG: No, it doesn't, sir.
And REMIX generally underpredicted all the
fluid downcomer temperatures. One of the reasons may
be that these locations -- the temperatures of the
centerline of the plume isn't located below the
downcomer injection but rather between two of the cold
legs.
That about wraps up what --
CHAIRMAN WALLIS: So the question is what
happens now. Are you going to keep working with
REMIX?
MR. YOUNG: No, sir. I don't believe --
I don't --
CHAIRMAN WALLIS: You --
MR. YOUNG: I don't believe that REMIX is
going to be able to consider all the physics involved
in determining the downcomer temperature transients.
CHAIRMAN WALLIS: So your conclusion is
that we should replace REMIX with something better?
MR. YOUNG: Yes, sir.
CHAIRMAN WALLIS: Okay.
MR. SHACK: Well, I mean if the -- if the
temperature differences are always as small as they
seem to be as in these experiments, I mean RELAP does
just as well, doesn't it?
MR. ROSENTHAL: I think the point is that
RELAP doesn't model the mixing. See, you wanted to
explore this aspect. And we back on the East Coast
tried REMIX also and we got offsets very, very similar
to their showing, and they were enough to be important
in the wrong direction.
We would have to have incorporated it into
the uncertainty analysis that we passed on to the
fracture mechanics.
And you do it -- you know, we did it
because it's cheap and fast, and whatnot. But I think
we'll have to make a decision with what we do REMIX
now.
And the point was that you couldn't -- I
mean Theo put it together to solve a problem in the
mid-'80s. He just -- you didn't have the options you
have today.
MR. YOUNG: We'll see in comparison to the
CFD Codes that the calculated temperatures are not in
the same accuracy of the CFD Codes.
Thank you.
CHAIRMAN WALLIS: Well, we have one more
minute.
Anyone have anything more to say? NRC
want to say anything more?
MR. BESSETTE: Well, I guess, you know, --
CHAIRMAN WALLIS: One minute.
(Laughter.)
MR. SHACK: Forty seconds.
MR. BESSETTE: You know, of course we
wanted to run REMIX because this was a code we
developed for PTS -- for the purpose of PTS analysis.
So, if nothing else, somebody along the
line would have said, well, why don't you run REMIX.
So we wanted to run REMIX against actual data and just
to see how it does.
But like we said, you know, when we
developed REMIX in 1985 we couldn't do CFD analysis,
but now we can.
MR. KRESS: And was -- REMIX was fitted to
real data back in '85. What this is telling me is
this data gives different results than the '85 data,
presuming the REMIX --
MR. SHACK: The Creare one-fifth test,
yeah.
MR. BESSETTE: Well, yeah. But, see,
REMIX was always run against the fluid-mixing
experiments, which were designed to look kind of like
REMIX. They incorporated only that part of the system
that REMIX models, so the cold leg and downcomer.
So that the experiments, the configuration
of the experiments matched the part of the system that
REMIX models. So maybe that helped REMIX to model
those experiments.
And this system is more of a -- you have
more of an integral system even when we're running
these sort of separate effects kind of tests where
we're just looking at the -- trying to focus on the
HPI injection and the plumes.
MR. YOUNG: And I'd like to mention is --
MR. REYES: Get to a mic here, Eric.
MR. YOUNG: Oh, I'm sorry. This is Eric
Young. I'd like to mention that the facilities that
REMIX was validated against didn't include any core
region and were simplified plume geometries, a planar
plume, and a rectangular duct, and didn't include any
of the endless geometry or heat from the core region
that we're seeing here. So it's not able to do that.
CHAIRMAN WALLIS: So I will declare a
break and we will remix here at quarter to 11:00.
(Recess taken from 10:30 a.m. to 10:45
a.m.)
STAR-CD AND CREARE HALF SCALE BENCHMARK CALCULATIONS
MR. HAUGH: I'll dim the lights just
slightly so you can better see the animations and
everything in the presentation.
I'm doing a presentation on STAR-CD and
the Creare half-scale PTS test facility benchmark
calculations. My name is Brandon Haugh. I was here
earlier.
The objectives of this presentation are to
introduce the STAR-CD CFD Code and basically just a
preliminary of what it is; a description of the test
facility that was compared; a description of the tests
that I ran and calculated; a description of the model,
the computational maps that I generated; and a
comparison of the results from calculation with actual
data from the test facility.
And I will make some conclusions about how
it compared.
CHAIRMAN WALLIS: Do you remember when the
Creare tests were run?
MR. HAUGH: These half-scale tests were
run, I believe, in 1987.
CHAIRMAN WALLIS: '87?
MR. HAUGH: Yeah.
Okay. Well, the objectives here are to
benchmark STAR-CD; provide insights into the CFD Code
operation to help with the APEX-CE simulations; and
also to establish a learning curve for the STAR-CD
Code.
CFD Codes are tricky to run, and so
there's a lot of things that you learn along the way.
The STAR-CD CFD Code is from a
computational fluid dynamics code. The acronym "STAR"
stands for simulation of turbulent flow in arbitrary
regions.
The code consists of three major
components: A preprocessor/postprocessor called
Prostar, which is where you build a mesh and it sets
up the problem.
CHAIRMAN WALLIS: Does it have automatic
grid generation also to stuff that makes it easy?
MR. HAUGH: It does, yeah. It makes it
easier. It's never easy, the CFE, but there's a
package called ISM CFD, which will allow you to take
a 3D like ProE-generated models and insert them and it
will generate the mesh automatically.
For the purpose of this study we're kind
of -- for the entire learning curve, I was using the
built-in mesh generating tools and built it by hand.
There's the analysis package called STAR
which is a Fortran Base Code that runs the problem and
does the calculations.
And there's also a parallel computing
interface called Pro-HPC, which is important in CFD
because you need a lot of computational horsepower.
CHAIRMAN WALLIS: Is this one that
converges on the answer; it does all kinds of
iterations?
MR. HAUGH: Yes, it does a lot of internal
sweeps --
CHAIRMAN WALLIS: And tells you what the
residuals are or --
MR. HAUGH: -- for every iteration. Like
it does sweeps --
CHAIRMAN WALLIS: Sometimes in these
problems it doesn't converge. I mean if the plumes
are wandering around all over the place, it won't
converge on any answer at all.
MR. HAUGH: Yeah. And it will let you
know if it diverges. It produces an error file, and
it will let you determine -- it will help you
determine what parameters you might have set up
incorrectly, or if your timed step was off, things of
that nature, yes.
The code has the capability of handling
many types of fluid flow, dispersed flow, and chemical
reactions, compressible flow, moving meshes, all kinds
of things.
The Creare half-scale test facility, which
ran their tests in 1987, was built to not model any
particular PWR but to be flexible with interchangeable
cold legs and injections, to be able to do different
PWRs.
The configuration used in the MAY-105 and
-106 tests is displayed here to the right. On the
next slide I will give you some characteristic
dimensions since they didn't show up too well. But
you can kind of see it's a planar downcomer, a cold
leg that's horizontal, top injection with a loop seal
and a pump simulator.
CHAIRMAN WALLIS: Because it's a downcomer
with walls on the side, that's what restricts the
plume, doesn't it?
MR. HAUGH: Exactly. Yeah. So it will
make a difference. It's different than an integral
facility or a full-scale plant.
Some of the characteristic dimensions from
the test facility, the cold leg inside diameter was
14.3 inches. HPI inside diameter, 4.5 inches.
Downcomer width, 63.7 inches. The gap 5.4. Thermal
shield thickness, one and a half inches. There was a
thermal shield in place, which also had a height of
100 -- let's see -- I forgot to put in -- of 95
inches. And vessel wall thickness for the core side
and the vessel side of 2.75 inches, and that was
carbon steel.
Now the MAY-105 test --
CHAIRMAN WALLIS: I think that was scaled
so that it was big enough so that it was thermally
thick and it didn't have to be any thicker.
MR. HAUGH: Exactly. Yeah. So you still
had -- you had a conduction-limited period and 8 d
back -- 8 d back period.
The test that was ran for the MAY-105 was
a stagnant loop, 462 k or 189 c. That's 372 degrees
Fahrenheit.
It had a constant HPI injection flow of
5.17 ten to the minus third meters cubed per second,
which is about 1.37 gallons per minute; at 14.2
degrees C, which is about around 70 degrees F.
The test duration was --
CHAIRMAN WALLIS: That's pretty low
velocity.
MR. HAUGH: Yeah. It was still pretty low
velocity. It was actually kind of comparable to the
APEX test. The velocity for the injection ended up
being about half a meter per second or about 1.6 feet
per second. So not terribly high, but actually I
think it will help compare to the APEX facility.
And test duration was a little over 2,000
seconds.
For the STAR-CD model a computational grid
was generated using the built-in tools in STAR-CD.
The resulting grid consisted of a little over 200,000
fluid cells and 60,000 solid cells representing the
steel.
I'll show here, this is what the
computational grid looks like, the red representing
the fluid cells and then the yellow representing the
steel. Sorry you can't see the refinement better.
It's kind of hard to get it on a PowerPoint slide, but
you can see that the downcomer -- actually you
probably can -- is far more refined than the loop seal
or the cold leg, because I wanted to try to capture
the plume behavior very well.
You can see the thermal shield in place.
And I only inserted steel on the core side and vessel
side of the downcomer since that was comprised of most
of the volume.
CHAIRMAN WALLIS: Now your nozzle here,
now their nozzle was more characteristic of a PWR, it
wasn't just a pipe stuck in with a sharp corner?
MR. HAUGH: Exactly. Yeah, it has the
gradient of a typical nozzle.
It's still a sharp edge. I didn't
incorporate a smooth edge, so there's some difference
there.
MR. SCHROCK: And you put your boundary
condition for the injection back up in the pipe there
somewhere?
MR. HAUGH: Yeah.
MR. SCHROCK: Is that what that shows?
MR. HAUGH: Yeah. Right at the top of the
pipe here is where the boundary condition for the
injection of the HPI was.
There was also a boundary condition here,
but the velocity was specified as zero, so it did the
correct pressure, back-flow stuff.
Now the outlet is different. You'll
notice the lower plenum isn't the same as the test
facilities. Theirs kind of came out here and was more
-- it was triangular.
Due to the built-in mesh-entering tools in
STAR-CD, that wasn't really a possible geometry to
replicate. So I tried to preserve volumes as best as
possible. And along with the curvature right here to
promote the mixing.
MR. SCHROCK: And this grid is cylindrical
in the downcomer?
MR. HAUGH: It's -- no, it's planar.
MR. SCHROCK: Planar.
MR. HAUGH: Yeah, which is what the test
facility was, too.
MR. SCHROCK: Oh, yeah. That was an
unwrapped in it, yeah.
MR. HAUGH: Yeah. And so the outlet is
actually right here. You can't see it. It's actually
more characteristic. It's right here. It consists of
six cell faces which comprise within three percent the
same area as the outlet of the standpipe on the Creare
facility.
CHAIRMAN WALLIS: It's a constant pressure
with a hydrostatic term or something there?
MR. HAUGH: Yeah. It's just -- it's
treated as an outlet so it won't let it drain. Just
kind of -- it's the same pressure as what the fluid is
in the lo- --
CHAIRMAN WALLIS: And there's no --
there's no problem with reverse flow. The problem
with outlets is if you can reverse flow it everything
gets confused.
MR. HAUGH: The code will get confused if
you have that, which I initially had, but I adjusted
some parameters and no longer see that in my runs.
But, yeah, there will be some problems with that.
Here's the volume comparison. I'm pretty
close in most things except for the pump simulator,
and that was user error. This was the first time I
was running CFD and the first mesh I had ever
generated. And so I was concentrating on the cold
legs and the downcomer and basically the lower plenum
and loop seal. And for some reason I lost my mind on
the pump simulator, and that's a large part of the
mixing volume there.
So you will be able to see the slight
discrepancies in the comparisons I'll show later, but
--
CHAIRMAN WALLIS: So you're going to fix
that?
MR. HAUGH: Yeah, I will. I was in the
process of it. But due to the geometry and how the
grid was, it became more time-consuming, so I couldn't
get it done in time, but I will have better results
for you.
The STAR-CD model inputs, you get to
establish some parameters for the model to be able to
run the problem, such as a turbulence model.
I chose to use a high Reynold's number,
k-epsilon model but, as Dr. Wallis pointed out
earlier, I didn't have a lot of experience in
turbulence modeling, so it might not predict the
correct interface in the stratified cold leg, so there
might be some errors associated with that.
There are several different turbulence
models that can be applied and any input that you can
provide would be very valuable.
The density was treated just as a function
of temperature, not pressure. It's basically a
constant pressure system so it's isobaric, just
function of the thermal expansion coefficient.
I used a time step of a quarter second.
I ran 4,280 iterations for about half of the tests.
Basically at that point in the test the facility was
almost all the way cooled down and so we capture most
of the important transient I think in this first part.
I can run the rest of it, I was just under a time
crunch --
CHAIRMAN WALLIS: My experience of CFD is
it's pretty grid dependent.
MR. HAUGH: Yeah. Yeah, exactly. Your
time step is related to your mesh size. And so this
produced sufficient convergence for me that I felt the
accuracy was okay.
I also used only an upwards difference
numerical discreditization scheme. And so it's just
a single-step first-order accurate method.
CHAIRMAN WALLIS: You're supposed to try
sort of good refinement in various areas --
MR. HAUGH: Exactly.
CHAIRMAN WALLIS: -- and also to see if it
makes any difference.
MR. HAUGH: Yeah. There's some gradients
around the thermal shield that probably need to be
refined, but as the first time in CFD, this is kind of
my best effort at this point.
MR. SHACK: What were your run times like?
MR. BOEHNERT: The next page.
MR. SHACK: Oh, never mind.
MR. HAUGH: Yeah, I'm getting there.
It ran for 7.7 days on -- I ran it in
parallel, so two processors. It's a Sun Blade 1000
work station, 750-megahertz processors, and one gig of
RAM. It's a UNIX terminal.
The results I'm going to present here are
some of the cold-leg stratification predictions, some
downcomer temperatures. I have some animations
showing the plume activity on both sides of the
downcom- -- over both sides of the thermal shield, so
the core side and the vessel side. It's pretty
interesting.
And just a couple of snapshots of some of
the convection circulation patterns around the base of
the thermal shield.
For my cold-leg stratification comparison,
there is a cold-leg raking in the Creare test facility
9.1 inches after injection towards the vessel. The
rake is centerlined, the spacing being about 1.43
inches between thermocouples. And they're labeled 1
through 10 starting at the bottom.
This is the greater representation at that
same location in STAR-CD, so you can see that my grid
isn't quite encapsulating every thermocouple location.
Sometimes two thermocouples will fall in one cell.
And also that I didn't have cells right on
the centerline, so in my comparisons I used both
cells. Since they were pretty close in temperature
they'll both be on the box.
And I'll show you, I compared position 1,
position 10, and position 4, which is right here,
which is this set of cells.
For position 1, the bottom of the cold
leg, I seemed to predict that pretty well in STAR-CD.
It looks like it may be slightly under the average of
-- read by the Creare facility, but the mesh was also
pretty course, so I feel that it did a fairly good job
representing that.
The next location of position 4, which is
slightly below centerline, I still did pretty well.
I even captured the step phenomenon that they captured
in their data for the mixing --
CHAIRMAN WALLIS: The wiggles are the
data, aren't they? The wiggly curve is the data.
MR. HAUGH: Yeah. The wiggly curve is the
data, due to the splashing in the interface and stuff.
And in my model, probably do the turbulence model, I
don't see that kind of --
CHAIRMAN WALLIS: Now you didn't fudge
anything? I mean you didn't change dials on the
model, or anything? This is just straightforward
using whatever's in the code?
MR. HAUGH: Yeah. I just specified a
mixing length --
CHAIRMAN WALLIS: You didn't have to tune
anything or --
MR. HAUGH: -- and a turbulent energy and
that was all.
There's lots of knobs you can turn, but it
kind of -- I think the kind of idea was, well, let's
see just straight out of the box what kind of -- well,
how good will it do. And it seems to do okay here,
but you notice the next plot, this is the top of the
cold leg, I seemed to not predict this very well.
Initially I do, but later I dip down
pretty good. It's as much as almost 50 degrees
Fahrenheit, so it's probably due to the mixing in my
turbulence model in the cold leg.
And possibly the cell on the top of my
pipe is quite large, and I didn't include steel on my
cold leg. So there might be some conduction there
affecting that. But I don't think it would bring it
up quite so much. So there's a little bit of
investigation that needs to be done into why I was so
far off, considering the other two locations seemed to
be pretty good. I just kind of still in the process
of figuring it out.
For the downcomer temperatures I created
an animation of the model to visualize the plume
activity, which is a nice thing about CFD is it will
give you some visualization which a lot of other types
of codes can't.
The first animation is via the vessel
side, which is the side that's important for
pressurized thermal shock and the second is the core
side.
CHAIRMAN WALLIS: Is there any other
option in this k-epsilon model you might use?
MR. HAUGH: Well, I can specify different
turbulent energies and mixing lengths. And I can also
--
CHAIRMAN WALLIS: But you can't account
for stratification, though, can you?
MR. HAUGH: There's different turbulence
model I might be able to use.
CHAIRMAN WALLIS: Does it have effective
thermal stratification on the turbulence itself? I
don't think --
MR. HAUGH: No, it doesn't incorporate
that.
CHAIRMAN WALLIS: It probably doesn't.
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: That's well known to be
a problem.
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: All the CFD people will
tell you it's a problem, --
MR. HAUGH: Yeah, but the --
CHAIRMAN WALLIS: -- but they won't give
you much of a solution to it.
MR. HAUGH: Yeah. They're good at that.
They say, oh, it can do anything, but maybe it won't
do that, but they won't tell you why.
So -- yeah, so there's probably some work.
The company that distributes this in the United States
has been pretty good at working with us for support
and stuff. And they're always interested in other
things they could add to their code.
CHAIRMAN WALLIS: Do you mind if I show
this to Creare?
MR. HAUGH: No, I don't mind at all.
Yeah, that would be fine.
So I'll start the animation here. I
recorded data every five seconds. And so to play the
whole test it's playing rather quickly. But you can
see that the plume activity on that vessel side isn't
terribly significant in terms of the gradient on the
side here. I mean really as the test goes on and it
cools down, it's maybe 10 to 15 degrees Kelvin in the
scale.
You can kind of see -- you can also see
that the loop seal riser and the top of the downcomer
don't participate in the mixing volume, which is
expected.
CHAIRMAN WALLIS: A real illustration of
sort of thermal waves are clearly different from the
fluid flow direction.
MR. HAUGH: Yeah, exactly. It's --
CHAIRMAN WALLIS: And nothing's moving up
in the downcomer.
MR. HAUGH: Yeah, so it's kind of neat.
CHAIRMAN WALLIS: This is distorted. Have
you made the downcomer extra fat in order to show
what's happening there?
MR. HAUGH: Well, I twisted it.
CHAIRMAN WALLIS: You twisted it? Ah,
okay.
MR. HAUGH: Yeah. So it's --
CHAIRMAN WALLIS: So we're looking at
different --
MR. HAUGH: -- a slightly angled so you
can see the --
CHAIRMAN WALLIS: Okay.
MR. HAUGH: -- the cold leg and the
downcomer.
Yeah. And these lines here represent the
top and bottom of the thermal shield.
What was interesting to see is in the next
animation that the thermal shield actually plays a big
role. It seems that the plume comes out of the cold
leg and hits basically right in the middle of the top
of the thermal shield and kind of splits and washes
back and forth. But it also has enough momentum that
it seems to -- you'll see right here -- it impinged
definitely more on the core side as the plume -- much
more significant --
CHAIRMAN WALLIS: Oh, yeah, it goes a long
way down there. Doesn't it?
MR. HAUGH: Yeah. But this is on the
core-barrel side, where in the PTS phenomena we don't
really care, so --
CHAIRMAN WALLIS: Hey, it looks like some
sort of a dancer there.
MR. HAUGH: Yeah, but it is. It's still
only like 20, 30 degrees Fahrenheit -- I mean Kelvin.
But it is.
CHAIRMAN WALLIS: But that -- look at
that, I mean that looks like a plume that's getting
narrow at the bottom.
MR. HAUGH: Yeah. It is. It's not --
CHAIRMAN WALLIS: Accelerate.
MR. HAUGH: -- what you'd conventionally
think of the rising. So it could be the CFD model and
the turbulence or --
MR. WACHS: So it's a temperature probe --
that's just the peak temperature in the center is
getting narrower.
MR. HAUGH: Yeah.
MR. WACHS: It's actually just sporadic --
CHAIRMAN WALLIS: Right. That's right.
We're just looking at temperature.
MR. HAUGH: Yeah. When you come over this
afternoon, I have much more information available.
Like I can show you velocity distribution of the
plume. And it's actually -- it does widen.
CHAIRMAN WALLIS: I think Walt Disney
could really put that to music.
(Laughter.)
MR. HAUGH: We could make millions. Yeah,
so it's pretty interesting.
Now for a comparison in the plots of the
actual data, this is a map of some of the
thermocouples in the Creare downcomer. They did have
more in this row, but these are just the ones I chose
for comparison, to not overdo you with too many plots.
I'm going to be comparing rows 5 and 9
just for -- to keep it brief.
In this plot I've included both -- there's
thermocouples on both sides of the thermal shield in
the center of the downcomer gap in that region. So
I've presented both the vessel and the core side.
I think -- probably in black and white
it's not very easy to see on the paper. It's a little
easier in color here. The Creare is the blue and the
pink. This is the row 5, column 7. So this is
centerline below the cold leg. And then the STAR-CD
data is the black and the red.
But you can see for the core side that
STAR-CD seems to do pretty well. It even predicts
some of the plume behavior. And due to the fact that
I recorded data every five seconds, it's not going to
get all the plume behavior, so just keep that in mind.
But it seems to do a reasonably well job.
At the bottom here I'm slightly below, and that could
be due to my mixing volume being slightly off. But on
the vessel side I seem to be underpredicting what
Creare does. I get the plume that he did pretty well,
but as the test goes on, you'll see I'm below what the
actual data presents. But I mean it's really not too
bad. It's maybe --
CHAIRMAN WALLIS: Which would -- which
would indicate there's really more mixing than you're
predicting?
MR. HAUGH: That's kind of what I'm kind
of leaning towards. Or it could be the mixing volume,
and so my cool -- cool-down rate is a little bit
faster than what it was in the real facility.
I'm working on producing these in a
nondimensional format with the mixing volume or the
mixing time on the bottom. And so I'm hoping that
will clash the data a little tighter, but we'll have
to wait and see.
This is pretty much the bottom of the
downcomer, or really close, centerline as well. And
you'll see that on the core side I'm pretty much right
on until the end. And of course that comes back up,
but I kind of stay down.
And on the vessel side I'm pretty much
below the whole time.
The order of magnitude is not too bad.
It's maybe 20 degrees Fahrenheit.
CHAIRMAN WALLIS: This is really quite
good for CFD and the new problem --
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: -- without any kind of
tuning or --
MR. HAUGH: Yeah, without turning the
knobs or anything.
CHAIRMAN WALLIS: -- tweaking.
MR. HAUGH: I mean relatively it's pretty
good.
I mean in terms of the overall like a
power plant operation, this is a much simplified
model, with the planar geometry, stagnant loop, and
just the injection. So --
CHAIRMAN WALLIS: And you've still got
that piece that you modeled wrong, too?
MR. HAUGH: Exactly. So when I correct
that I'm hoping the data will be much more agreeable,
I mean given the fact that it's so good now, hopefully
it should get better.
And here's the well-mixed temperature for
the inlet of the core, which is -- which would be the
standpipe temperature in the Creare facility, which is
just the outlet of my model.
CHAIRMAN WALLIS: This is just a heat
balance, isn't it?
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: So you think you should
get that pretty --
MR. HAUGH: Well, if my volumes are off,
--
CHAIRMAN WALLIS: Yeah.
MR. HAUGH: -- you can see it right there.
So I think when I nondimensionalize it and fix my
models, I think it should be just about right on. So
it seems to be meeting that pretty well.
Okay. Now for the velocities on the
thermal shield, this is just kind of interesting to
note. I'll kind of explain it a little more.
Now this is the opposite side of the plume
entry. So the plum, the cold leg was not in the
center of their planar section downcomer, it was more
to one side. So on the opposite side of that at the
bottom of the thermal shield, you noticed we have a
convection pattern so we have upflow. So there's a
cell that built inside of the downcomer.
Now if you'll look at the other side where
the plume is we see the downflow, but there's still a
slight -- a swirl on the core side right here. But I
just thought it was interesting to point out. And the
velocity magnitudes are given here. So it has decayed
and slowed down a little bit because of the density
difference.
This is also at 60 seconds into that
transfer --
CHAIRMAN WALLIS: That could be a bit
awkward because somewhere in that -- that swirl, on
the edges of it, you've got stagnation points, --
MR. HAUGH: Yeah. Well, actually I --
CHAIRMAN WALLIS: -- and most-heat
transfer correlations would --
MR. HAUGH: -- this is more looking this
way. And I can show you later that these velocities
are actually impinging the wall.
CHAIRMAN WALLIS: They are?
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: They can't go through
it. So somewhat there you've got a stagnation point
--
MR. HAUGH: Yeah.
CHAIRMAN WALLIS: -- because you know it
devised. And so I think your heat transfer and your
CFD would probably predict there's no heat transfer
there or very, very low H.
MR. HAUGH: Yeah, it probably will. I do
have heat-flux data that I didn't present here because
it's very new, but I can show you later if you want to
see.
Just to mention it, my heat fluxes that
the CFD predicted were pretty -- on order of two in
some locations greater than what the Creare reported.
But their measurement was kind of, well, here's the
thermocouple in the middle of the fluid. Here's the
thermocouple at the wall, and they guessed the heat
flux.
CHAIRMAN WALLIS: Now this is an
instantaneous picture. If you did it 20 seconds later
it would probably look quite a lot different.
MR. HAUGH: Yeah, it would look completely
different given the transient case. I can show you an
animation back in the lab if you'd like to see what it
looks like. It's pretty interesting.
Some preliminary conclusions I made is
STAR-CD is a benchmark against the Creare data for a
first-case, rudimentary analysis.
The well-mixed temperatures were slightly
underpredicted, but that's most likely due to my
model's mixing volume being 6.7 percent less than the
Creare facility. So I apologize for that. That's my
error.
Predictions of plume temperatures in the
downcomer compared reasonably well with the data.
Considering that I didn't do much adjustments or
anything, I felt it was reasonably good.
CHAIRMAN WALLIS: How long did it take you
to get to the point where you actually could --
MR. HAUGH: I started doing this last
summer, and so about a year. And I'm getting
reasonable results.
CHAIRMAN WALLIS: And you ran it last week
or --
MR. HAUGH: I ran it and finished a month
ago.
CHAIRMAN WALLIS: A month ago, okay.
MR. HAUGH: Yeah. And I've been running
jobs throughout and just -- just kind of -- initially
I had my turbulence wrong because I was learning. And
so as you learn, you go, oh, well, that wasn't very
smart, and so you rerun it. And then it takes some
time --
CHAIRMAN WALLIS: Did you learn on this or
did you learn by running a lot of other problems
first?
MR. HAUGH: Well, I kind of learned on
this just seeing, well, that doesn't look right. And
then with information from Adapco and Dan Wachs
helping with the input, people with more experience
than myself helped me do a better job at establishing
the correct parameters for the model.
And it's probably a good thing to point
out the CFD is largely an experience based kind of
usage of the code. The more you learn about -- I mean
because you need to know about a good background in
turbulence and things like that to be able to apply
phenomena in models to correct circumstances. So --
so I learned how to use STAR-CD using parallel
processing, which was kind of interesting on its own.
Parallel computing is kind of relatively
complicated. I'd like to acknowledge the computer
support staff at the College of Engineering helping
out with that. They did a great job.
But it was obviously -- the parallel
computing helped me speed up the problem quite a bit.
And the benchmark calculations aided in
the selection of turbulence models, the selection of
the fusion links, and turbulent intensities for
preliminary runs. You can see that things weren't
quite as they were supposed to be.
And that's the end of my presentation.
CHAIRMAN WALLIS: If you look at various
weather patterns you can see how stratification kills
the mixing and you get distinct layers and they go a
long way.
MR. HAUGH: Yeah. And something
interesting that I was thinking might be able to help,
this was kind of after the fact, is the Oceanography
Department here does extensive meteorological research
on vast supercomputers. And they might be able to
give better insight into what models they use to
represent stratification.
CHAIRMAN WALLIS: Yes. yes.
MR. HAUGH: Yeah. Well, thank you.
CHAIRMAN WALLIS: Thank you very much.
3-D CFD MODEL OF THE APEX-CE TEST FACILITY
MR. WACHS: Okay. I guess I get to finish
up here. We're going to talk about the 3-D CFD model
we used for the APEX-CE facility. And on this work I
worked with both Eric Young and Brandon. And we got
quite a bit of help from the Adapco folks for the U.S.
distributors for computational fluid dynamics.
I'm a grad from Oregon State. I'm now at
Argonne, so I have to do a little shift in gears from
sodium pool fast reactors back to water reactors. I
think I can make the switch.
First off, I'll talk a little bit about
what our goals were with this model.
We wanted to explore the potential of CFD
modeling, to treat some of these individual phenomena
that we were seeing in the reactor out there that we
didn't think codes like RELAP and REMIX were going to
be able to capture.
And to do that we chose one particular
test, we chose the OSU-CE-3 test to try to model and
see what we could come up with. After I talk about
that a little bit I'll talk about our particular model
and the components we included in the model and things
we did with that.
I will look at a couple of the phenomena
that we were able to observe in both the model and the
APEX facility, including stratification of the cold
legs, comparison of the core inlet temperature, the
downcomer temperature profiles. And then we'll try to
extrapolate to some -- some data that they used in the
Creare facility with some models they used to look at
plume velocities and heat-transfer coefficients. I
will summarize that.
And then after that we'll speak just a
little bit about some of the lessons we learned on
CFD.
On this particular test the objective was
to look at cold-leg stratification and the downcomer
profiles. And that's why it was a good one for us to
attack with CFD. In the particular test the reactor
coolant pumps were turned off and natural circulation
was driven entirely by -- by core decay heat.
We did a couple of different tests, and
this is the one Dr. Reyes showed where he had five or
six different stratification plots.
We chose to go with the 200-kilowatt case
because that gave us the greatest amount of
stratification experimentally, and we wanted to see if
we could capture that.
Describing the model, it's -- the
objective of the model is to capture all this thermal
hydraulic behavior in just the cold legs and the
downcomer. But in order to do that we had to include
some of the peripheral pieces of the facility for
either reasons of convenient boundary conditions or we
thought that it might be participating in behavior to
a certain extent.
In particular, for inlet conditions we
needed to include the loop seal and the HPSI. We
couldn't inject directly into the cold leg; you
weren't going to get a real good profile.
And we had to assume adiabatic walls
outside of the model.
CHAIRMAN WALLIS: You had to assume that?
MR. WACHS: Well, you don't have to, but
you can -- you can specify heat fluxes on the walls.
But you're not necessarily going to know those
beforehand.
CHAIRMAN WALLIS: Unless you model them.
Didn't we learn from some of the earlier presentations
that the heat transfer from the wall matters?
MR. WACHS: Oh, yeah. Oh, yeah, we have
the wall included and we have the outer vessel wall
included. But on the outer vessel wall.
CHAIRMAN WALLIS: Oh, the outer of the
outer vessel wall altogether.
MR. WACHS: So convecting off to the
environment. Right. So it's insulated, but there's
still a certain amount of convective loss.
CHAIRMAN WALLIS: Oh, I see. I thought
you meant --
MR. WACHS: Yeah.
CHAIRMAN WALLIS: -- the inside, too. So,
no. No, your -- where it matters you actually modeled
the heat transfer.
MR. WACHS: Yeah -- well, yes and no,
because we included the outer wall but not the inner
wall, and that shows up as a problem later on. That's
something that would certainly be added in the future.
The boundary conditions we treated in this
particular test. I think that's a typo on the initial
temperature. I think it was closer to 400 Fahrenheit.
But the loop-flow rates, in cold leg 3 we had 14
gallons per minute in the HPS -- or through the loop,
and 12 in the other cold leg. The HPSI lines were
about a half gallon per minute and a gallon per
minute.
And these were extracted directly from the
test facility. So we ran a test, got our boundary
condition and applied it to this -- to this model.
Now here's a picture of the model. You
can see the two cold legs on each side. We have a
loop seal attached to each. The HPSI lines are coming
in at an angle on the horizontal plane.
We have the downcomer. We have the full
region of the downcomer. And in the center we had to
include pieces of -- well, we had to include the core
region in order to keep away from numerical problems
with changing directions and backflow, which Dr.
Wallis mentioned earlier.
And we also have a half of the core vessel
overlayed on top of that as solid cells.
One thing that's important to note,
though. Since we were initially just treating the
internal core region as a stop gap for numerical
problems, it's adiabatic, those cells are not
connected. Okay, so there's no communication
temperature-wise between those two regions.
Now here's a closer look at what we used
and where we assigned boundary conditions. Here's the
cold leg, and we assigned a boundary condition right
at the edge of the HPSI line directly from our
facility. The HPSI line is long enough so we can get
a pretty full developed flow and get a good idea of
what the mixing may be like. And the steam generator
inlet boundary condition was the loop-flow rate.
One thing you might note is that we did
try to maintain as much of the geometry as we could
reasonably include, so the injection nozzle does have
the tapered approach. There are sharp angles on the
inside, though. You start rounding them, the cells
get to be really difficult to draw and maintain.
One thing you might notice that right at
the HPSI injection, it's really hard to see the mesh
density in here. We doubled -- well, actually we
quadrupled the density of the mesh at HPSI injection.
And I think that that was a good first step, but in
the end I don't think it was enough. We should have
done some more. The injection region was larger than
that.
CHAIRMAN WALLIS: Maybe you need to get a
smaller mesh just where -- where the jets coming in
and then --
MR. WACHS: Right. Right, and that's what
we tried to accomplish there, but the jet I think ends
up being a little bit longer.
CHAIRMAN WALLIS: Does this STAR-CD enable
you to refine the jet in places where you say have big
velocity gradients, or something?
MR. WACHS: Yeah, you can do adapted
meshing.
CHAIRMAN WALLIS: Automatically, yeah.
MR. WACHS: And one of the problems -- and
you had mentioned this problem earlier in that one of
the things you always want to do with the CFD is you
want to prove that it's mesh independent and just keep
increasing the density of the mesh till you see it
doesn't change.
CHAIRMAN WALLIS: So the answer doesn't
change, right.
MR. WACHS: But in this case it took 10
days to run. And when we --
CHAIRMAN WALLIS: So you'd either run out
of money or the answer doesn't change.
(Laughter.)
MR. WACHS: Yeah, that's right.
CHAIRMAN WALLIS: Or time, right.
MR. WACHS: And --
MR. ROSENTHAL: Ten days to run on your
laptop?
MR. WACHS: No, on -- on a four processor
Sun. So it's -- we have substantially machinery we're
running it on.
MR. YOUNG: A four-parallel processor.
MR. WACHS: Yeah, right.
So -- but absolutely, that's something
that needs to be done. And I think that we need to
address our computational ability in order to be able
to do those things. Until we can do that it's hard to
really say we have the right answer. But we're
working on pieces of that. I think that's coming
along.
So in this case we wanted to extract some
of this data from the model after it had been run this
transient over -- I think we went between 3- and 400
seconds. I can't remember, it was like 600 time
steps.
(Brief discussion held beyond the range of
the microphone.)
MR. WACHS: Okay. So we wanted to extract
some of the data from the model and compare it to what
we saw with the APEX test facility in the thermocouple
rakes. And this next plot -- and these are the -- and
these are the cells that would coincide with those
particular temperature thermocouples.
All you could see are the red lines are
our model. And the black lines are from the APEX
facility. You could see that the -- well, we are
getting thermal stratification. It doesn't match up
as well as we'd like, unfortunately. And I think
that's driven by the fact that we didn't include
enough cells in the region to really show a fully
developed mixing region. Maybe the size of the cells
were smaller than the -- or larger than the scale of
the mixing phenomena, or something.
Anyway, that's an area where increasing
the grid density may be an effect. In fact, Eric did
some tests --
MR. SHACK: And you're at k-epsilon again?
MR. WACHS: I'm sorry. Go ahead.
MR. SHACK: You're at k-epsilon
turbulence?
MR. WACHS: Yeah, we used k-epsilon in
this case. You know like we said before, we started
with a default, the ones that generally worked the
best, to see what would happen.
And it would be nice to do a parametric
study on several other of the models they have
available and see how it works. And I think it's been
shown time and time again that changing the turbulence
model changes your results, and you want to find the
one that works best for your case.
CHAIRMAN WALLIS: There seem to be more of
these calculations or six of them, then there are five
of the APEX data. So is something missing in the APEX
data there?
MR. WACHS: I think some of the upper ones
may be overlaying on top of each other.
CHAIRMAN WALLIS: That close? They do
have wiggles. It would be unusual for the wiggles to
overlay. It looks as if there are five APEX groups
here and six -- is there one reason the -- it looks as
if there's an APEX missing in the middle or there's an
extra CD where there isn't an APEX measurement or
something.
MR. WACHS: I had to change -- I changed
the colors on them. I may have grabbed the one and
changed it incorrectly to a wrong color and just had
it disappear. I can -- I can replot that for you this
afternoon if you'd like.
MR. YOUNG: One thing is -- this is Eric
Young. One of the things we'd like to mention is we
did do a refinement on the cold leg with the loop seal
geometry and everything and reran it for a stagnant-
loop condition and achieved very accurate results in
this cold-leg stratification or the temperature
grading across the cold leg for the same geometry.
MR. WACHS: Right. And by just looking at
the cold leg we were able to get the cell small enough
that you could run the test relatively quickly, in the
course of a day easily.
MR. YOUNG: Two days to make that right.
MR. WACHS: Two days? Yeah. So that
lends to that effect, but I -- yeah.
CHAIRMAN WALLIS: So it stratifies more in
reality than you predict?
MR. WACHS: Yeah. We saw a greater degree
of stratification. I think the model showed more
mixing than there really was.
CHAIRMAN WALLIS: And that's what you'd
expect.
MR. WACHS: Yeah, right.
I go onto the next one. Looking at the
core inlet temperatures, the APEX facility shows
warmer inlet temperatures than we saw in the STAR-CD
model. And one of the reasons that we're postulating
for that is that we didn't include communication with
the downcomer or the core barrel, okay.
And had we included thermal communication
between those two -- two pieces, we would expect the
temperature for the STAR-CD model to shift up. And
whether it would reach and match, we don't know, but
I think it was partially to move it in the right
direction.
CHAIRMAN WALLIS: You should get this
fairly well. This is an energy balancing --
MR. WACHS: Yeah, I think so too. I think
so too -- well, --
CHAIRMAN WALLIS: It's not so sensitive --
MR. WACHS: Yeah, at that point --
CHAIRMAN WALLIS: -- to the plumes and all
that stuff.
MR. WACHS: -- it's all well mixed, that's
right.
CHAIRMAN WALLIS: It's all mixed up, isn't
it, by now?
MR. WACHS: Yeah. I would think that
that's a fair estimate.
MR. YOUNG: One thing that needs to be
mentioned about the core inlet temperature is that the
location that you choose to actually compare this, the
-- you can just choose one node or cell at the core
region. Now if you went and chose a cell at a
different location and with the plume interaction and
hitting it, you're going to get different
temperatures.
CHAIRMAN WALLIS: Well, the plume doesn't
go that far.
MR. WACHS: So there's still some mixing
behavior going on in the lower plenum.
MR. YOUNG: Yeah. So there is some mixing
behavior going on in the lower plenum. So the choice
of the cell location with the temperature --
CHAIRMAN WALLIS: I thought it was well
mixed long before that.
MR. YOUNG: There's -- it's well mixed,
yes, sir. It is well mixed. But have certain stunts
of plume and interaction where it will fold down into
the region. And large recirculation zones will occur
and you will get kind of a stunt of water go down.
The temperatures between those stunts
aren't that much, but it will change the accuracy
slightly.
MR. WACHS: Yeah. If you look at the
temperature distribution it's only 10-degrees
difference. It doesn't change a whole lot. So if you
get just a mild recirculation where maybe this half of
the downcomer is cool and it's falling in, it's still
going to be displacing some hot fluid that's sitting
in there, so it's a dynamic behavior.
CHAIRMAN WALLIS: Well, then you'd expect
more wiggles perhaps in the data, wouldn't you?
MR. WACHS: Yeah. Well, the facility, I
don't think it's got some of these -- I think we're
overpredicting some of the convective behavior --
CHAIRMAN WALLIS: Something is wrong about
the heat input or something here, because you're off
by a large amount in the temperature change at the
end, by almost a factor of 2. So it looks as if some
source of heat or something's missing in the model.
MR. WACHS: Oh, yeah, absolutely. That's
-- I agree. That's where the --
CHAIRMAN WALLIS: That's what you said at
the beginning, I think.
MR. WACHS: The core barrel effect is I
think important then.
CHAIRMAN WALLIS: You could estimate that,
couldn't you?
MR. WACHS: What's that?
CHAIRMAN WALLIS: Can't you estimate that?
Do some --
MR. WACHS: Off the top of my head, no.
CHAIRMAN WALLIS: Some quick -- yeah, back
of the envelope, transient heat transfer.
MR. WACHS: I don't know what the total
mass is of that, so --
CHAIRMAN WALLIS: But you can find that
out if you --
MR. WACHS: Yeah.
CHAIRMAN WALLIS: Someone will tell you
its shape and size, and you can just do a calculation.
MR. WACHS: Oh, yeah, sure. If you know
what the injection flow rates in the initial -- yeah,
I agree.
Looking at the downcomer temperatures, if
you look at 1.3 cold-leg diameters below and we tried
to compare the flows, the STAR-CD calculation is -- is
a bit lower than what we're --
CHAIRMAN WALLIS: It seems to suddenly go
wrong at one point and never recover.
MR. WACHS: This one does?
CHAIRMAN WALLIS: Yeah. It seems to go
wrong at about 80 seconds and then it never comes back
to --
MR. WACHS: To lift back up to that --
CHAIRMAN WALLIS: Something happened at 80
seconds to get it wrong.
MR. WACHS: Well, that's where injection
begins, where you start to see the cold fluid falling
in.
CHAIRMAN WALLIS: Oh, okay.
MR. WACHS: And I think that this is
clearly the worst agreement between the downcomer
ones, this one right below. And I think we see some
dipping behavior in the --
CHAIRMAN WALLIS: Well, isn't this the
business of it hitting -- going across and hitting the
inner wall?
MR. WACHS: The thermocouple, yeah. So
that's kind of the problem with these -- with
comparing these analyses. You'd have to grab a single
point out of the facility and hope that your phenomena
you're looking for crosses it. And it may or may not.
Because realistically we can't expect with this kind
of behavior STAR-CD to exactly match what -- what's
happening at the facility because it's somewhat
unstable behavior.
CHAIRMAN WALLIS: So you don't have many
nodes across the downcomer, do you?
MR. WACHS: Radially or axially?
CHAIRMAN WALLIS: Radially?
MR. WACHS: Or azimuthally? Not very
many.
CHAIRMAN WALLIS: No. So --
MR. WACHS: But you saw with Kent Abel's
work, when he had the plot in Excel with the single
data points, those were our points. And you could see
the plume moving from place to place.
MR. YOUNG: Dr. Wallis, the number of
nodal locations crossed down by the gap is 8.
MR. WACHS: Oh, okay.
CHAIRMAN WALLIS: Oh, so you should be
able to pick up the difference between the inside and
the outside?
MR. WACHS: Oh, yeah. Oh, yeah.
Definitely. And I have -- I'll talk about that a
little later.
CHAIRMAN WALLIS: You're actually -- this
is STAR-CD predicted at the location on the outside --
MR. WACHS: Yes.
CHAIRMAN WALLIS: -- where the
thermocouple is. Oh, so that's not the explanation
then.
MR. WACHS: On the next slide you're
looking at two diameters below. It seems to do a
little bit better. Kind of crosses through the middle
of all the wiggles.
And at three diameters it's still pretty
good.
MR. SCHROCK: Your calculation seems to
smear out some oscillations in the actual --
MR. WACHS: Yeah. I -- that's definitely
true. I don't think that it's -- well, again we're
looking at the effect of turbulence and some of these
eddies and whether the code will be able to capture
that, I don't know whether it will or not. Apparently
it looks like it doesn't catch it as well as reality,
but we still get -- the mean behavior is similar.
So at four diameters it looks pretty
similar also. We're in the right ballpark at the very
least.
This is a vector plot of the plume
velocity at a cross section in the downcomer. So
we're in the middle of a downcomer. We've peeled off
the layers. And you can see where these plumes are
going. They're obviously merging together and
interacting into a single plume. And we actually see
a convection cell around --
CHAIRMAN WALLIS: And those colors are
what? They're velocities?
MR. WACHS: It didn't come across very
good. The darker -- or the redder colors are faster
velocities in the Z direction.
CHAIRMAN WALLIS: So it gets faster as it
goes down?
MR. WACHS: Well, that's kind of -- one of
the problems with this is that we have several layers
to choose from. And by choosing this middle layer it
seems to be fastest at the bottom, where if you choose
the layer -- chose a layer closer to the inside of the
core barrel it would be higher up.
CHAIRMAN WALLIS: Well, that's very funny
because we were told that the plumes dissipate after
about 4Ds and here they are going faster at the
bottom.
MR. WACHS: Well, you have to consider the
scale.
CHAIRMAN WALLIS: I'm not sure any --
there's any excuse for it, is there? It just seems to
be different altogether.
MR. WACHS: Yeah. Well, we actually see
some temperature -- well, this is just a model, too.
And we haven't really had an opportunity to compare
velocities from the model to that of the facility.
CHAIRMAN WALLIS: Yeah, but the STAR-CD
did so well in the more detailed analysis we just saw.
MR. WACHS: In the Creare facility?
CHAIRMAN WALLIS: Yeah.
MR. WACHS: Yeah. And that's a smaller
facility, too. And I think that the geometry --
CHAIRMAN WALLIS: Well, this is kind of
surprising. You're saying that you get these big
velocities at the bottom of the downcomer?
MR. WACHS: Well, in this particular
slice, --
CHAIRMAN WALLIS: And someone else --
MR. WACHS: -- that's where the peak
velocities are.
CHAIRMAN WALLIS: -- is telling us that
the plumes die --
MR. WACHS: I'm sorry. Go ahead.
CHAIRMAN WALLIS: -- four ds below the
injection point.
MR. WACHS: Say it again. I'm sorry, I
didn't get it.
CHAIRMAN WALLIS: I'm just trying to
reconcile it. We were told earlier that the plumes
essentially die and everything is well mixed up to
about 4 ds.
MR. WOODS: Well, there's --
CHAIRMAN WALLIS: And here we've got to
these plunging plumes which are more intense at the
bottom than the top.
MR. WACHS: Yeah. I guess I'm not really
willing to say that that's the most intense region.
CHAIRMAN WALLIS: Well, that's what that
red flash showed us.
MR. WACHS: Right, in this particular
slice. I think if we took a slice in a different
location, it would change that -- that look.
CHAIRMAN WALLIS: Could you back up and
show us that again, that red --
MR. WACHS: Yeah, I can show you that.
MR. REYES: I think the other thing is
that this test we had a plume with cold flow.
CHAIRMAN WALLIS: Yes.
MR. REYES: So you have downcomer flow due
to the cold-leg flow. And I think we're seeing kind
of a mixing due to that.
CHAIRMAN WALLIS: Yes. But still you've
got these intense velocities.
Could you go back? And if you could tell
us what the red magnitude is compared with the red
there, the background?
MR. WACHS: Okay, it's a radial.
CHAIRMAN WALLIS: We saw a red thing in
the middle there, then. That velocity is very much --
well, that yellow patch, how -- what's that.
MR. HAUGH: That's about a point -- looks
like .21.
MR. WACHS: Yeah, that's about right.
CHAIRMAN WALLIS: What's the average
velocity?
MR. HAUGH: The average velocity looks to
be about .1 --
MR. KRESS: Two, big red.
MR. HAUGH: -- .1 --
CHAIRMAN WALLIS: Go back to that red one,
there. Go back to that big red smudge there. Another
one, there's another one. Well, it happened twice.
MR. WACHS: Yeah.
CHAIRMAN WALLIS: So it's not -- it's not
just erratic.
MR. WACHS: Yeah. It's about two-tenths
of a meter per second.
CHAIRMAN WALLIS: Compared with an average
of --
MR. HAUGH: Of .15, it looks like.
MR. WACHS: Right, I think that's about
there.
CHAIRMAN WALLIS: So it's another big
deal.
MR. WACHS: Yeah. So it's -- yeah, I
would agree. I think that's a strange behavior from
a model.
But, again, this is -- before we were
talking about some stagnant cases. And it's possible
that -- I don't know -- maybe we're missing the plume
with the 8 d thermocouples. I think it will be a
little clearer when we get to later on we'll see --
CHAIRMAN WALLIS: Well, I thought that
some of the conclusion we seemed to be coming to from
the previous presentation was that we should replace
REMIX with CFD, because CFD does better and models
more things, catch they data better.
MR. WACHS: Yeah.
CHAIRMAN WALLIS: And this seems to be
showing that CFD can also predict things which may be
--
MR. WACHS: Oh, absolutely. There's no
question about that.
CHAIRMAN WALLIS: -- we could lose faith
in its ability.
MR. WACHS: Yeah. Well, we should. You
should -- yeah, I am sure you guys know that CFD is
not a black box. These codes are not black-box codes.
CHAIRMAN WALLIS: Well, I think we have to
make an assessment of the probability of CFD giving
good enough answers.
MR. WACHS: Right. I think what -- it
just hits me. In my personal opinion I think what we
would need to do is if we were ever going to
incorporate the CFD Code is we would have to develop
a mature code that worked well for a particular set of
geometries.
And we would want to understand that code
and feel comfortable with the results it gave us
before we actually went out and applied it to a
general case, or to another case. I don't really
think we're at a mature stage on this code yet. I
think that it still lies on the young state of this
particular model.
CHAIRMAN WALLIS: Well, the remarkable
thing was Brandon did something which was as immature
as possible. In fact, he hadn't tuned anything.
MR. WACHS: Right.
CHAIRMAN WALLIS: He just took something
out of the box and raised it. And it seemed to work
very nicely.
MR. WACHS: Oh, yeah, right. I agree.
It's hit and miss.
CHAIRMAN WALLIS: And it's the same code
that you are running here.
MR. WACHS: It's the same code, but it's
a different model, so --
CHAIRMAN WALLIS: Different person.
MR. WACHS: Well, I would -- well, I don't
know.
CHAIRMAN WALLIS: So it's used as a
dependent.
MR. WACHS: He helped.
Yeah, -- no, it is used to depend, that's
definitely true. And does the model you apply capture
the behavior you're looking for? So on this case we
have loop flow. In Brandon's case we didn't have loop
flow.
CHAIRMAN WALLIS: I know what is. It's
that he's a student. You're working for ANL. Isn't
it?
MR. WACHS: I'm a student, too, still.
CHAIRMAN WALLIS: Oh, okay. I thought you
were --
MR. WACHS: I just transferred to a new
location.
CHAIRMAN WALLIS: It's an ANL effect, is
it?
MR. WACHS: Yes. So -- but, yeah, I
definitely think that -- you know, we talked about the
size of Brandon's model. He used about 2,000 nodes,
and we're about 750,000 nodes. And that may have some
impact as well.
MR. REYES: This is a more complicated
case, though.
MR. WACHS: Yeah. There's --
MR. REYES: The other case was a stagnant
injection -- injection into a stagnant region. And
this is a flowing case, but not only flowing, but some
asymmetric injection --
CHAIRMAN WALLIS: In a sense none of these
are explanations as to excuses --
MR. REYES: Right.
CHAIRMAN WALLIS: -- or possible
hypotheses and --
MR. REYES: Hypotheses as to why we don't.
CHAIRMAN WALLIS: So it would be
interesting to resolve this.
MR. REYES: Absolutely.
MR. WACHS: Sure. Because we came in, we
even stated a couple of times that it's an exploratory
issue. We're trying to see how well it will do. And
I think it's important to note that it worked really
well in one case and it's not working so well in the
other case. So there's a certain level of confidence
that you should try to extract from that.
The next thing I tried to --
CHAIRMAN WALLIS: You're not seeing --
you're not seeing this stratification effect that we
heard about before, are you?
MR. WACHS: In the downcomer?
CHAIRMAN WALLIS: You are, you're getting
more of it than was predicted. I thought earlier you
were predicting -- your red code showed more
stratification.
MR. WACHS: That's only in the cold leg.
That was the cold-leg behavior.
CHAIRMAN WALLIS: Oh, that was in the cold
leg.
MR. WACHS: Yeah.
CHAIRMAN WALLIS: But not in the
downcomer?
MR. WACHS: Not in the downcomer.
Actually, the downcomer, it seemed to work pretty
well. Other than that first node right below, it --
no -- was so-so. But the other node seemed to work
pretty well.
Then from these velocity profiles we went
in and tried to extract the peak velocity for the
plume in order to compare to some of the work that
they did at the Creare facility in modeling with what
they saw experimentally. So we took off this peak
velocity and got a Reynold's number for the flow.
And that what they do at Creare, they use
these Reynold's numbers. Well, they picked out their
peak velocity and tried to compare it to -- they tried
to calculate a peak velocity and compare it to their
actual facility. And we tried to do that same thing
here.
In our case, we -- looking at each
individual cold leg as an independent plume generator,
and using this model that the Creare people used to
come up with a maximum plume velocity, you know, in
one case we would have two plumes. We got the lower
of the two curves here. So we got two independent
plumes of moderate velocity.
When you combine those two plumes and you
see they have the same strength, they have the QHPI, it
goes together, you get a larger plume as you'd expect.
And this larger plume, the model predicted, seems to
match better with what we saw in the model, whatever
that's worth.
CHAIRMAN WALLIS: I'm not quite sure what
I'm looking at here, these points way down the
left-hand side there.
MR. WACHS: Yeah. I'll get it for you.
Yeah. Now what do I have to do to roll.
These points down here, these are from the model. And
what this is showing is it's showing the --
CHAIRMAN WALLIS: Model's warming up, or
something?
MR. WACHS: Yeah. It's getting started.
The plume is forming. So the velocities are low as
it's forming, initially. In terms of the model that
they, the Creare people, which we are trying to apply,
it doesn't treat that. It just said the plume was
there and it's performing in a certain way.
CHAIRMAN WALLIS: Well, once it gets going
this isn't all that bad, then?
MR. WACHS: No. Yeah, right. And it
developed, you know, a pseudo-developed plume. It
seemed to match okay. Now that's just a check to see
a guess what they got. And it's kind of interesting
that you have to combine the plumes in order to get
that similar behavior.
CHAIRMAN WALLIS: Where do you recall this
maximum plume velocity?
MR. WACHS: Well, as you saw in the --
CHAIRMAN WALLIS: Is it that red flash, is
the maximum plume?
MR. WACHS: Yeah, basically that's what
you have to do, because the plume is always moving
around. You look for the hot spot, and you say, "Oh,
that's the peak velocity at this particular point in
time."
And so you have a group of cells which you
grab, and you say the velocity is something like that.
And that's one of the real challenges with comparing
this type of model data to real data. It's...
Now this is a look at the downcomer fluid
temperature. In this case, I superimposed to the mesh
over the top. And they look better out in the sun
than they did on the presentation. Again, this is a
mid-plane temperature in the middle of the downcomer.
CHAIRMAN WALLIS: So here we've got a
plume which is going way down.
MR. WACHS: Well, the temperature
gradients here, it's like 475 to 468.
CHAIRMAN WALLIS: Can we freeze it? Let's
see -- no, go back. Go back one or two.
MR. WACHS: See, I don't know if we have
to let it finish before we can --
CHAIRMAN WALLIS: Go back -- you have to
go through?
MR. HAUGH: Yeah, and it will finish,
then.
MR. WACHS: I think we do. It's a movie
file in place.
CHAIRMAN WALLIS: You have to start again,
or something?
MR. HAUGH: Yes.
CHAIRMAN WALLIS: Because there was a time
there where you had a plume which seemed to bring --
again, it's difficult to see the colors, but there's
a yellow plume that seemed to go all the way down.
MR. HAUGH: Yes, exactly.
MR. WACHS: Um-hum.
MR. SHACK: Yeah, but the whole
temperature range is what, 468 to 474?
MR. WACHS: Yeah, it's not a very big
temperature range, right.
CHAIRMAN WALLIS: So it's very little.
MR. WACHS: So, you know, if you see it
cooling you're going to see profiles. It's not a very
strong plume.
CHAIRMAN WALLIS: So what's the
temperature of the fluid coming out of the cold leg?
MR. WACHS: Yeah, like I said, this is a
mid-plane, so it's jumping over the plane that we are
looking at.
CHAIRMAN WALLIS: But what's it coming in
at?
MR. WACHS: What's it coming in at; the
temperature rise of the plume?
CHAIRMAN WALLIS: Yeah, what's the
temperature when it comes out of the cold leg?
MR. WACHS: I would assume it was coming
in at about what the thermal -- or the cold leg, cold
stream, the same.
CHAIRMAN WALLIS: Is that the 468, or
something, whatever is at the bottom there?
MR. WACHS: I would say that it would be
a little bit cooler than that. I think there's
certain amount of mixing that has gone on before it
gets back to this plane.
It's coming in through these holes here
(indicating). These are a couple of cells that were
part of a different part of the model.
CHAIRMAN WALLIS: It may be there is
something which is just distorted here. It may be
that there's a big temperature change at the top,
which you're not really seeing.
MR. WACHS: Right.
CHAIRMAN WALLIS: And then there is a
survival of a plume at the bottom, because there is
some mixing, even though it's stratified. And we are
just focusing on that because that's all we can see.
MR. WACHS: Yeah, I agree.
CHAIRMAN WALLIS: I don't know.
MR. WACHS: That's probably true.
CHAIRMAN WALLIS: Hard to tell.
MR. WACHS: I think that, yeah, a majority
of the mixing is going to go on in this initial
falling area where it's impinging on the far wall.
CHAIRMAN WALLIS: It doesn't mean to say
that there's no mixing velocities below that. There
seem to be a lot of mixing velocities below that. Bi
because it's all about the same temperature it doesn't
matter.
MR. WACHS: Right.
CHAIRMAN WALLIS: So it seems to be that
you have to really separate out your idea about what
the velocities are doing from what the temperatures
are doing.
MR. WACHS: Right. Yes. It's the post-
processing. There's a lot of data to sift through and
choose which ones you want.
Did you want to see that again?
MR. SHACK: Was it possible to refine the
mesh right up in that region, right at the nozzle?
MR. WACHS: Oh yeah, absolutely.
MR. SHACK: I mean it --
MR. WACHS: That's a thing you can do.
MR. SHACK: And that seems to be where the
action is.
MR. WACHS: Yeah, that's one of the things
that you'll want to beforehand. We'll talk of little
bit about that when we get to some of the summary, how
we would do that.
Now this is a movie of the vessel wall
temperature. So the temperature differences here are
on the order of four degrees from top to bottom on the
range. You can see they seem to mimic the temperature
profiles in the fluid pretty well. But the gradients
are really small, what it's expecting to see.
CHAIRMAN WALLIS: Again, this is
emphasizing what maybe a rather small changes at the
bottom of the annulus.
MR. WACHS: Yes.
CHAIRMAN WALLIS: And not really showing
you that the mixing is occurring at the top.
MR. WACHS: Yeah. Well, this is actually
the steel, the first node of steel. So showing the
overall cooldown of that body.
CHAIRMAN WALLIS: Because it's surprising
the steel is changing temperatures so rapidly.
MR. WACHS: Well, it's not changing much.
CHAIRMAN WALLIS: It's not?
MR. WACHS: No. This is the -- the full-
scale from red to black is less than four degrees.
CHAIRMAN WALLIS: Okay. That's part of
it. You've magnified it.
MR. WACHS: Yeah. That's essentially what
it does.
MR. SHACK: If you look at it close
enough, there's always big differences.
MR. WACHS: Yeah. And, you know, if I
hadn't done that it would be red. You know, you don't
really see any of the behavior.
CHAIRMAN WALLIS: If you go long enough,
he's going to write OSU in the annulus.
(Laughter.)
MR. WACHS: Yeah. I'll have to play that
on the scoreboard at the football game.
The next step that we wanted to try to
accomplish there was to try to extract some heat
transfer coefficients and see what we could look at.
In order to do that --
CHAIRMAN WALLIS: Now who is this guy
Newton? Is that a reputed reference?
MR. WACHS: Yeah, that's by name-dropping.
So we want to extract it. In order to
come up with this we have to extract a heat flux from
the problem, and we have to find some way to
demonstrate the delta t. And heat flux isn't too bad.
You just take a delta t over the reactor vessel wall.
And we know the thermal conductivity of that
particular material so we can extract the heat flux.
And again you'll want to do that at the --
or I did that at the area of the peak velocity, a
plume velocity we chose.
Then you just plug it in and choose an
ambience temperature. I chose the temperature at the
mid-plane, just to have something to work with. And
you can translate that into a Nusselt number.
In the Creare work they compared their
Nusselt numbers that they calculated and measured to
-- so they Spelter equation. They didn't get good
agreement, either. And we tend to not see -- be off
by about -- well, we're around a hundred and they were
around two hundred.
MR. SCHROCK: Well, that's kind of a
casual determination of the representative fluid
temperature.
MR. WACHS: Yeah, it is.
MR. SCHROCK: You need something more
definitive than that.
MR. WACHS: Absolutely, I agree. And one
of the hard things with doing that is you get a
nonuniform temperature profile. Ideally you would
like to choose a mixed mean temperature of the plume.
And how to do that is not trivial.
CHAIRMAN WALLIS: Maybe this is one of the
problems with REMIX. And REMIX has to do things like
get an average heat transfer coefficient in this sort
of way.
MR. WACHS: Right.
CHAIRMAN WALLIS: And this is kind of
indicating that that is not a very good
representation.
MR. WACHS: Right. It's a constructive
parameter. Extracting a heat transfer coefficient is
not -- you know, this kind of code isn't intended to
do that kind of thing.
CHAIRMAN WALLIS: Your CFD has to do
something about predicting the wall heat transfer
coefficient.
MR. WACHS: Not really. It's just -- it's
looking at nodes.
CHAIRMAN WALLIS: Yes, it does. You can't
just -- you know, but that doesn't give you a
coefficient of the wall. There's got to be some model
that goes from the velocities in the nodes to a heat
transfer coefficient at the wall.
MR. WACHS: Well, use a strict convection
of and conduction between the two.
CHAIRMAN WALLIS: It uses some kind of a
law of the wall. It uses some kind of a model of
what's happening in the boundary there.
MR. WACHS: Yeah. It does use the law of
the wall. That's incorporated into there.
CHAIRMAN WALLIS: Right. It's not clear
that that applies when these plumes are doing what
they are doing here.
MR. WACHS: Right. I agree.
CHAIRMAN WALLIS: So that you've got the
same problem with CFD, but there's a different level
than that.
MR. SCHROCK: Dave says just use infinity.
Is that universal, then?
MR. WACHS: Between all the codes?
I'm sorry, I didn't get that.
MR. BESSETTE: Well, I think what we've
shown is that if you're within a factor of 2, then
that's plenty good enough. If so, if you choose --
that's why -- like in REMIX, you know, in REMIX you
input the heat transfer coefficient.
So you can just choose something like a
thousand-watt square meter degree k. And if you're
off by a factor of 2, it doesn't matter. It's 500 or
2000.
CHAIRMAN WALLIS: Don't use an infinity of
it because then you'll find that five different
calculations will give you infinite changes in the
step and you'll be in real trouble.
MR. KRESS: I'm surprised that the data is
below the Dittus-Boelter.
MR. WACHS: But in this next slide we'll
talk about that a little bit, how that works, and why
that is.
CHAIRMAN WALLIS: Or a lot.
MR. KRESS: Yeah. I would have thought it
would have been above it.
MR. WACHS: Just a little bit on the
Dittus-Boelter equation. It's based a fully-developed
flow, and that's not really the case we are looking
at. In our case --
MR. KRESS: Yeah, you've got an interest
region.
MR. WACHS: -- if you look here, one of
these is --
MR. KRESS: How do you calculate the
Reynold's number?
MR. WACHS: -- the one on the left --
CHAIRMAN WALLIS: Well, that's the whole
point.
MR. KRESS: I think that's the issue
there.
MR. WACHS: On the left side here we've
got an axial slice of the temperature profile. On the
right, we've got an access slice of the velocity
profile in the z direction.
If you look in the upper-right quadrant
there you can see where the cold plume is located at.
And, you know, that's about five degrees. And this is
about mid-plane. You can see that it's attached to
the inner wall. It's clearly not running down the
center of the vessel -- of the annulus.
And you can see on the velocity profile
that the plume is moving with that cold stream just
like you would expect.
MR. SHACK: Where am I at in z again?
MR. WACHS: This is right by the mid-plane
of the downcomer. So I don't know. What's the total
depth of the downcomer?
MR. HAUGH: Total diameter is 86 inches.
MR. WACHS: It's like 71 inches, but I
don't -- so it's probably around 35 inches. So that's
about 10 diameters from the upper-vessel head, not
from the downcomer, or from of the cold leg injection
plane.
So we can see that the actual velocity
next to the outer wall is relatively small. And thus
you'd expect the heat transfer coefficients of the
model was -- would you extrapolate from the model
would be smaller than what you'd get with a
fully-developed flow with the cold plume running down
the middle as opposed to the far side. So the
effective diameter for the Reynold's number is
probably different.
CHAIRMAN WALLIS: These two plots
correlate pretty well, don't they?
MR. WACHS: Yeah. Oh, I think so, with
the velocity and the --
CHAIRMAN WALLIS: Which I think means that
they're calculating the heat transfer coefficient from
the velocity. And that gives you the temperature. So
you would expect them to correlate pretty well.
MR. WACHS: With the Dittus-Boelter
equation?
CHAIRMAN WALLIS: Well, whatever model
they have in CFD for the heat transfer coefficient.
MR. WACHS: Oh, oh, I didn't use the --
CHAIRMAN WALLIS: It's probably -- is
probably --
MR. WACHS: The CFD model didn't extract
the h coefficient. I extracted that from the data.
CHAIRMAN WALLIS: Oh, you extracted that?
MR. WACHS: Yeah.
CHAIRMAN WALLIS: You extracted that.
MR. WACHS: Yeah. You have to do some
special things beforehand in order for it to calculate
an h value. You have to --
CHAIRMAN WALLIS: I think CFD would make
h roughly proportional to v.
MR. HAUGH: Yeah. I did that in my Creare
model on the last run. What you have to do is between
the solid and the fluid cells you insert a wall
boundary. And it's just zero resistance. That's just
a point for the code to monitor heat flux. And then
from the heat flux you can provide a mix mean
temperature, and it will give you a heat transfer
coefficient.
MR. WACHS: Right. Yeah, it will
calculate the y plus value and then try to extract the
heat transfer coefficient. But it's something you
have to do before you run the model really early on in
the development.
Just some of the conclusions we could
make: Like we said before the nodalization was a
little bit coarse in the cold leg, and I think we
could get better results if we were to densen that up.
And as a part of that we need to show great
independence.
The downcomer temperatures seem to be in
pretty good agreement.
The core inlet temperatures were mildly
underpredicted. And, as we postulated before, I think
that's due to the omission of the core barrel.
Now in terms of phenomena base we do see
plume interactions, which is something that we think
we've seen in the facility from the data that we have
available. And those interactions should be affecting
the plume velocities.
In addition, the plume doesn't necessarily
run right down the middle of the downcomer. It moves
in its radial location from inside to outside,
primarily down the outside. Generally that affects
the agreement with some of the standard convective
coefficient -- heat transfer coefficient models.
We also really need to run some more runs
and tweak the model a little bit to include some of
the physics that we want to make it match the data a
little bit better, or to model the same problem,
essentially. Essentially, we are not treating the
same set of problems. And we also need to show that
the cell density is appropriate.
MR. SCHROCK: So your heat transfer --
MR. WACHS: There should be one more on
lessons learned.
Any questions?
MR. SCHROCK: Your Nusselt number
comparison is even worse when you recognize the
Dittus-Boelter equation is for the length average in
a long tube, fully-developed conditions. And you're
applying it in developing conditions.
MR. WACHS: Right, I agree. They don't
really treat the same problem. The only reason I put
that on there is because the Creare people put that on
there when they looked at their data. I just wanted
to compare how well our model was working against how
well the model they used.
This should be another one lessons learned
that you've got in there.
MR. HAUGH: What's the name?
All right, I got it.
MR. WACHS: Also I'll talk a little bit
about the lessons we learned from this stuff, with
Brandon -- it took Brandon and Eric and I working on
it for the last year. There's some significant things
to take with us.
On the CFD methodology there's some
distinct steps to that you need to go through in
creating a model. And each one of those steps has got
its own set of problems and difficulties that you can
face.
The first step is you really need to
define your problem well. You need to know exactly
what behavior you're looking for. You need to know
exactly what your geometry is going to be like, where
you know inlet conditions and boundary conditions.
And so you need to be familiar with the
system before you come in. It's going to be really
difficult to get any reasonable results if you come
into a system fully blind on the physics.
The next step that you have to go through
is you really need to construct the mesh. You'd have
to construct the mesh. And the way you construct that
mesh needs to be attached to the way you want to model
the physics and the physics you want to capture.
Finally, you need to set up your problem.
So you need to choose your models, what physics do you
want to include, which turbulence parameters you want
to -- or turbulence models for you want to run. And
then you need to run the problem. And that goes into
the computational ability that you have in your
facility, how you go about doing that and what you can
do. And then post-processing presents its own set of
problems.
So things to consider, but you have to
understand the physics before you go in. It's not
going to be something that you treat as a black box
and you just take somebody who knows how to run
software, and you put them on it and try to run the
problem and expect to get anything that's worth
anything.
In terms of mesh building, one thing we
found -- that Brandon found -- was that he found it
easier to import the geometry from some CAD program
that was more suited to generating rough geometry.
In general, the CD or the STAR -- or the
CFD codes are not real good at generating initial
geometries. They're good at generating meshes, but
they're not good at initiating initial geometries.
You don't have the same amount of tools. And you can
be more efficient and more complete in representing a
geometry when you do that.
Now your mesh gradients need to match the
physical gradients of the phenomena. It's really
important to do that.
And a lot of times people think of CFD as
you put more nodes in, and you make time step smaller,
it's going to work better. And that's not always
going to be the case. You can get too small just as
well as you can get too large. So you need to be
careful about that as well.
In terms of running the codes, modeling
large --
CHAIRMAN WALLIS: So interesting. You can
refine the cells in local places but you have to run
everything at the same time, that you step --
MR. WACHS: Right. You have to rerun the
entire problem. Yeah, because there are all
communicating with each other.
CHAIRMAN WALLIS: You can't refine the
time in certain regions?
MR. WACHS: Well, yeah. I guess -- yeah,
I guess I'm thinking of time -- transient time scales.
CHAIRMAN WALLIS: I guess you could, but
I don't think they ever do that.
MR. WACHS: Do what, vary time?
CHAIRMAN WALLIS: Refine the time.
MR. WACHS: Oh, oh, yeah.
CHAIRMAN WALLIS: In certain regions you
run the very small time set only in this region, --
MR. WACHS: Oh, okay, yeah.
CHAIRMAN WALLIS: -- but the rest you're
on --
MR. WACHS: Yeah, you're right. You have
to vary the time globally. You can't vary it in time.
CHAIRMAN WALLIS: Globally but not
locally.
MR. WACHS: Yeah, you're right. At least
not at this time. I wouldn't be surprised if they
decided that they could do that, whether they could or
not.
The next thing, when you dealing have
large systems and have large numbers of nodes, you
need a lot of computing power to be able to treat
those models. That's where parallel computing really
came in useful for us. Had we not had the parallel
computing capability we'd still be running models
probably months ago. You know, it would be -- our
particular model with the APEX-CE facility took 10
days to run on a four processor parallel machine.
So you take that onto a single processor
and you're talking about order of months. So that's
a significant advantage.
Now another thing we found was that you
needed to have your computing platforms homogeneous.
You couldn't work with an HP and a Sun separately.
The code didn't like that. It wants to stay on the
same platform all the time.
And in terms of post-processing, one of
the main difficulties in comparing to experimental
data, even in a very well-instrumented facility like
ours, we are on the order of hundreds of data nodes,
okay, with the -- with the CFD we're on the order of
a hundred thousand locations.
So how you take this two-dimensional or
even three-dimensional behavior and try to benchmark
it against basically one point in time and really get
an idea of whether you're seeing the same behavior.
And that's difficult. I think that that's --
CHAIRMAN WALLIS: Which is always a
problem with code assessment --
MR. WACHS: Yeah.
CHAIRMAN WALLIS: -- with RELAP or
anything else. I mean you get a whole lot of data
points and a whole lot of predicted points and how you
assess the comparison between them --
MR. WACHS: Right. It's not trivial.
CHAIRMAN WALLIS: -- in other than some
sort of superficial way, or you just look at a few
pictures and say, "All right, it's good enough."
MR. WACHS: Right.
Let's see. Yeah. So we were talking
about multi-dimensional behavior. I think that's it.
That's -- because we've add sections.
CHAIRMAN WALLIS: So we are getting very
close to the end.
MR. WACHS: What's that?
CHAIRMAN WALLIS: We're getting very close
to the end here.
MR. WACHS: Yeah. Any questions on that
at all?
CHAIRMAN WALLIS: Any questions?
(No response.)
CHAIRMAN WALLIS: So, Jose, are you ready
to wind this up for us?
MR. REYES: I'm ready to do that.
MR. SCHROCK: So we have a new NRC
dilemma. When is enough CFD enough? Never.
MR. SHACK: When you run out of money.
SUMMARY AND REPORTING SCHEDULE
MR. REYES: As I went through this I
realized there's a lot to summarize. And I'll kind of
hit on some of the highlights that I think we saw as
far as observed phenomena and what we've learned in
the process.
In terms of separate-affects-type
behavior, we've taken a look at using this both
APEX-CE and the flow visualization group. We take a
look at a specific mixing behavior in the cold legs.
We looked at thermal stratification.
We found that the effect of the Weir wall
was very important. So that was something that was --
originally we didn't think that much about, but as we
went into the testing and we observed the results,
especially in the flow visualization group, we
realized that that was an important part.
The separate affects can be integrated now
with the integral system test, because we see that the
formation of a cold liquid plug in the loop seal can
affect integral system behavior. It affects
stagnation in the loop. So that's a very important
piece of our research, I think. And so we will
continue development in that area as far as analytical
models.
In terms of the integral system testing
one of the things we observed there, or some of the
key points that we observed for the integral system
test, were for all the small break LOCAs, the main
streamline breaks, and then the combination breaks,
what we observed with regards to the mixing in the
downcomer was that the plume basically within four
diameters was well mixed.
For cases where we had flowing cold legs
in conjunction with injection we also saw thermal
stratification. And that ranged -- the maximum we saw
was about 40 degrees F from the bottom of the
downcomer to the cold-leg location. So we were seeing
small temperature differences.
We recognized the unique design of the
facility we're scaled here to, the Palisades plant.
It does have a lower HPI than some of the other plants
that are out there today. And so we recognize that
aspect of it. And so we are seeing some of that
effect.
We generally tend to produce relatively
weak plumes which mix relatively quickly in the
downcomer. That's what we're seeing.
With regards to the two situations, the
stagnant injection in a stagnant media as opposed to
a cold flow. For a stagnant media case what happens
is that initially you see good-sized temperature
differences between the -- within one or two d, about
15 degree F difference between the ambient temperature
and the plume temperature.
But as time goes on you're -- in the
stagnant case you're mixing that volume, and it's
becoming more and more cold. And so you're seeing a
smaller delta t with time and a weakening of the plume
with time.
With regards to the situation we have of
flow in the cold legs, what we see there is a
possibility of having a prolonged contact with the
downcomer vessel, because you are resupplying that hot
fluid. And so you're maintaining a larger delta t for
a prolonged period of time.
So as long as you're feeding hot water
into the downcomer region, you're able to keep this
plume relatively strong because you're continuing to
inject.
However, we saw maximum case of about --
I think we saw between 30 and -- I think 30 to 35
degrees was the maximum we saw for the situations that
we were setting.
However, it does provide an opportunity
for a prolonged contact but it's fairly small,
temperature-wise.
With regards to our analytical
capabilities, we looked at RELAP5 Code, the Gamma
version. And for the main streamline breaks, the
prediction seemed to compare very well. There were a
few areas that we still need to look at.
In particular, the pressurizer liquid
level we saw didn't predict exactly as we measured in
the test.
And then there are some issues with
regards to break flow, for the small break LOCA that
we wanted check into also as far as comparisons there.
So there's still some work in that area
that needs to be done with regards to the integral
system modeling.
In general, we observed that RELAP
predicted the stagnations reasonably well as far as
the stagnation occurring. The mechanism was typically
either reverse heat transfer or steam generator tube
draining.
It could not take into account the effect
of the cold loop seal plug because of spillover over
the Weir wall. So we noticed that.
With regards to the separate-effects-type
modeling, we used to two codes. We used REMIX Code
and we used STAR-CD.
REMIX is a control volume lump parameter
type of a regional-mixing mode, which is significantly
simplified. It just addresses certain regions within
the mixing volume of a cold leg loop seal and
downcomer.
And we observed, in comparing to our data,
that it didn't -- it underpredicted basically all of
the temperature is in the downcomer that we were
looking at.
We also observed that with REMIX you can
include and you should include a section of the core
barrel to model the core stored-energy release or the
barrel stored-energy release into the fluid. And that
played an important part. And that effect was also
carried over, we saw, into the STAR-CD calculations.
For STAR-CD we used a -- we did two
calculations. We benchmarked the code first with the
Creare half-scale data. And we saw a very good
comparisons right out of the box with that.
And I think if you look at the APEX-CE the
temperature scaled was fairly tight. And so I think
the comparisons are actually better than they appeared
on the screen.
So if you go back to that again, I think
you'll see you're still within about plus or minus --
within 10 degrees of the actual measured. So go back
to that and look at the scale, and you'll see that
it's a little bit tighter than the picture tends to
reveal.
So we do see a better prediction than I
think was the image portrayed somewhat. Would we
think we are seeing some reasonable predictions with
the STAR-CD code.
There's more to learn with regards to the
turbulence models and what's being done there.
Overall, if you look at the project as a
whole and what we were trying to accomplish and where
we are right now, I think we've accomplished quite a
bit. And so I'm pleased with that.
We will be documenting all of this and
providing you with the Final Report. And we'd like to
make it fairly comprehensive so it's a standalone
document. So it will describe the entire project in
its entirety.
There's some additional work that's going
on right now. We are still looking at some
theoretical models for the plume.
We're also looking at a theoretical model
for a prediction of loop seal spillover.
We're also looking at some refinements
that have been done, I guess, already on the cold leg
to see if we can't -- using STAR-CD to predict better
comparisons of stratification by refining the cold leg
a bit more. So we have those.
And we also have one test which right now
we still need to specify a little bit better, Test
Number 13. And we're looking at discussions with NRC
as to what's the best way to portray that. That would
be basically a cold injection of HPI into essentially
a steam environment where you have a level in the
downcomer. So we'd like to take a look at that to
understand that better.
So that's kind of where we're at.
With that I'll talk about the reporting
schedule. The Scaling Analysis Report is completed.
We've submitted it to the NRC. And it's actually gone
through the technical review. And now it's going
through an editorial review. So we are waiting for
comments back from NRC Publications. Not comments
from technical but the publications folks to get back
with us and get us into the right format for the
report.
The Final Report, we are planning to --
we'll probably have several, at least one or two
drafts before the Final Report, which we'll provide to
NRC for their review. We hope to issue that by the
end of the year. That will include the following
information:
Review of the previous PTS research. That
will be a section in there.
Description of our test facilities.
An overview of Palisades operations, what
we learned as far as operations of the Palisades
plant.
The results for all of our tests. It will
include the RELAP5 comparisons. And that includes all
the comparisons. We didn't show you all of them.
We will include the REMIX and STAR-CD
final calculations.
And then, of course, we'll have a CD with
all of our data and all of our drawings for the
facility. So typically we can bundle that essentially
on three CDs. So you'll get a three-CD set. You can
pick up your copy at the door. So that's how we'll
document this project.
So I think it's a fairly comprehensive
piece of work. And hopefully you will be able to
benefit from this.
But I would like to express my thanks to
the Committee for being here. And it's always a
pleasure to get a review. And on behalf of our
students, who have worked very hard and I think they
themselves have learned quite a bit from this process,
I extend the thanks to the Committee and hope we can
interact with you again in the future.
CHAIRMAN WALLIS: Well, thank you.
I think we've enjoyed hearing from you and
your colleagues and students. You've given us a lot
to think about.
I think the Agenda calls for a
Subcommittee caucus. I think what I'd like to ask is
what the ACRS role will be in the future.
Usually we review things like this in the
context of our review of RES programs when this
happens once or twice a year. This will be very
useful input, I think, for that purpose.
But I don't think the whole ACRS is
specifically going to focus on this project at any
particular time, except in the context of PTS or some
safety issue when we will really get involved.
So I see this mostly as contributing to
our decisions we have to make about PTS, any
rulemaking, particularly, and any regulatory action
which the Agency is going to make.
MR. BESSETTE: Well, I think that's right.
I think this -- you know, the PTS work here is -- it's
very issue-oriented as opposed to generic, like -- so
it's different than when you review the code
development or something like code consolidation.
This feeds into the -- this was intended
to feed into the work on the thermal hydraulics
aspects of PTS and it was a key part of it.
CHAIRMAN WALLIS: Now down the road I can
see -- we've been reviewing vendor codes. As you
know, we've got three or four of them we're doing now.
And presumably down the road the NRC, or the
licensees, or somebody may wish to approve in some
formal fashion the use of CFD in various forms, just
the way that they approved the use of GE Code, or
Westinghouse Code, and so on.
So I can see down the road we may have to
face that question. And these sorts of results will
presumably be part of a review of that type. But
that's probably down the road somewhere.
I assume that what we are going to do is
we're going to go back and at the next full Committee
meeting we'll make a report as a Subcommittee, which
will be 15, 20 minutes, half-an-hour, summarizing what
we heard here, anything that the whole Committee needs
to know.
MR. ROSENTHAL: Yeah, let me just
reiterate. I mean, with respect to PTS, you know,
we've got a schedule. We'd like to continue on.
We're doing plant-specific RELAP calculations.
We're going to do some benchmark against
this facility, price the calcs we're doing, but based
on the work that they did with RELAP. But I would
expect that that would come out rather well.
We had some show-stopper questions in
terms of plumes and stagnation, which I'm encouraged
in terms of getting some answers. So we start PF- --
MR. CHOKSHI: The thermal hydraulics
condition in September.
MR. ROSENTHAL: In September. And then we
are due to come to the ACRS in September, also.
MR. CHOKSHI: In other words, the
methodological --
MR. ROSENTHAL: And what I think you've
asked us to do is spread out, you know, a
beginning-to-end for some sample case. So it would
involve three divisions in Research. And I don't
think we would belabor the thermal hydraulics very
much, given what the Subcommittee has heard.
MR. SHACK: Well, it seems to me you've
gotten lots of visual insights into thermal hydraulics
and PTS from this room.
But what are your next steps now in the
thermal hydraulics part of the PTS analysis?
MR. BESSETTE: Well, basically we've done
our analysis of Oconee, which is the B&W design.
We've run about 60 transients, call them scenarios, or
whatever, PTS, potentially PTS-significant transients
to map out the plant.
MR. SCHROCK: Using RELAP?
MR. BESSETTE: Using RELAP.
We've done selected transients as well
with TRAC. So we have a TRAC RELAP comparison. So
that work is done.
We've started our calculations on Beaver
Valley and Calvert Cliffs. Beaver Valley is a
Westinghouse 3 plan and Calvert Cliffs is a CE. And
we hope to be well along in those analyses by, let's
say, October, which is when the fracture mechanics
people need the results to run through FAVOR.
And so that's where we stand.
MR. SCHROCK: And they need a --
MR. SHACK: But what kind of calculations
are you going to do assure yourself that the plume
behavior you see in those other plants is comparable
to the plume behavior -- I mean, the conclusions here,
I think, are that, you know, the PFM guys don't have
to -- you know, they're golden. And, you know -- but
can you reach that conclusion generically yet?
MR. BESSETTE: Well, as far as -- you
know, as far as a downcomer goes, basically, you know,
you have your initial conditions, which is what enters
from the cold leg.
And we would expect to cover the range of
conditions that we could expect from -- you know, for
a B&W plant you've got a high-mixing region when HPI
comes in.
CE and Westinghouse is similar in that you
get the stratified flow coming into the downcomer.
The HPI flows are slightly higher in the
Westinghouse Plant and CE, but we'll, you know, we'll
cover that, that range.
CHAIRMAN WALLIS: So are you going to
accept Westinghouse CFD calculations, something like
what we saw here, which are actually, you know,
reflecting the conditions in their plume, which is
somewhat different than there.
Are you going to except those predictions
for temperature, and so on, as inputs into --
MR. BESSETTE: Well, actually, we aren't
even getting any submittals from Westinghouse.
CHAIRMAN WALLIS: No, but I mean, what are
you going to do about the plumes, then? Is someone
going to make a calculation of what the plumes are
doing in a different plant?
MR. SHACK: I mean, Jack mentioned you
were looking at REMIX, which doesn't look very promi-
-- very good here, I mean.
MR. ROSENTHAL: Well, you know, I always
-- first of all, we are not getting -- let me just
clear. We're not getting anything, any calcs from
Westinghouse. We're doing the Westinghouse calcs.
CHAIRMAN WALLIS: Well, I was just
hypothesizing.
MR. ROSENTHAL: Yeah. Now --
CHAIRMAN WALLIS: But somebody is
responsible for facing the issue of: Is there a plume
and how big is it, and what are their temperatures?
MR. ROSENTHAL: Well, I think of the
three, I am, but I'm going to enlist the two people to
my right.
But I always saw this as a question of we
would do some RELAP calculations, and we would have to
associate an uncertainty with those calculations.
And I always saw that it wouldn't be a
constant uncertainty. Dave has pointed out to me more
than once that when -- for those sequences in which
you have pumps running, for example, you have a very
well-mixed case. And the uncertainties ought to be
smaller than in the stagnant cases, et cetera.
So I always, at least in my head, thought
there would be that qualification on which sequences
and we'll know the associated probabilities of those
sequences.
Now, should it come to pass that we have
some critical cases, either in terms of extremes of
pressures or temperatures, with high-enough
probabilities that you care about, then we're going to
have to do some more homework.
And REMIX we could always run, and it's
cheap. And as long as we have sequences in which you
don't cool down very much, or the pressures aren't
very high, and they can live with 25 c, or 50 c
uncertainty, then I don't -- I'm not convinced that we
should do any more.
But as soon as I get less than twen- -- if
they start telling me that 25 c delta really matters
in some sequence, then I think we're going to have to
do some more work.
CHAIRMAN WALLIS: From what you've seen
here, the work that's being done here, by the time
that you can extrapolate to the Final Report, you
think the information in that Final Report is going to
be just what you need to make these decisions?
MR. BESSETTE: I think so.
CHAIRMAN WALLIS: You think so?
MR. BESSETTE: Yeah.
CHAIRMAN WALLIS: That sounds very good,
then, as an output from this research work.
MR. ROSENTHAL: Oh, yeah, it's very
encouraging.
CHAIRMAN WALLIS: So, Jose, you heard
that. I think that's a very good outcome. These are
getting to be more internal NRC-type discussions that
you're just listening to.
Does the Subcommittee have anything else
to say at this time, or shall I just wind things up by
complimenting all the speakers.
It's been very interesting. It's very,
very nice for us to see technical work done from
theory, and experiment, and asking questions, and
getting answers. It's been a good couple of days.
Thank you.
(Whereupon, the meeting was adjourned at
12:17 p.m. on July 18, 2001.)
Page Last Reviewed/Updated Tuesday, August 16, 2016