Sei sulla pagina 1di 11

Nuclear Engineering and Design 235 (2005) 25702580

The stability boundary for supercritical ow


in natural convection loops
Part I: H
2
O studies
V. Chatoorgoon

, A. Voodi, D. Fraser
Department of Mechanical Engineering, University of Manitoba, Winnipeg, Canada R3T 5V6
Received 13 December 2004; received in revised form 10 June 2005; accepted 10 June 2005
Abstract
Non-dimensional parameters governing supercritical ow instability boundary, derived in an earlier study, are now examined
through a numerical experiment comprising 94 simulated cases. Part I of this study reports on the stability of supercritical light
water in a natural-convection loop and conrms the validity of these non-dimensional parameters for stability predictions. Part
II reports on supercritical CO
2
and H
2
.
2005 Elsevier B.V. All rights reserved.
1. Introduction
Future reactor designs are seriously consider-
ing higher temperature and higher pressure primary
coolant conditions in a bid to boost overall plant ef-
ciency (Dimmick et al., 2002). This includes the possi-
bility of operating with supercritical conditions on the
primary side. The ow instability of uids at super-
critical conditions is therefore of considerable interest
for this new application. For a nuclear reactor, oper-
ating at the higher supercritical conditions can lead to
substantial improvement of the thermodynamic ef-
ciency of the plant. New reactor designs in Canada,
DOI of the original article:10.1016/j.nucengdes.2005.06.004.

Corresponding author.
E-mail address: chatoorg@cc.umanitoba.ca (V. Chatoorgoon).
USA, Japan and Russia are seriously considering this
alternative.
In the early 1960s, Harden and Boggs (1964)
and Walker and Harden (1964) reported an analytical
and experimental study of a closed natural-circulation
system with Freon 114 near the critical temper-
ature. The exit heater temperature was, at times,
near the critical temperature but did not exceed the
critical temperature. Cornelius and Parker (1965)
reported heat-transfer instabilities near the critical
point. Their experimental study involved Freon 114
and the test section outlet temperature also hovered
around the critical temperature without exceeding
it.
Zuber (1966) did an extensive reviewand the rst in-
depth analytical study of the various instability modes
of supercritical uid ow. He considered ow in a
0029-5493/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2005.06.003
V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580 2571
Nomenclature
A ow area (m
2
)
C
K
loss coefcient in momentum equation
(m
1
)
D hydraulic diameter (m)
f friction factor
G mass ux (kg/(m
2
s))
G
m
maximum mass ux (kg/(m
2
s))
G
s
mass ux at the stability boundary
(kg/(m
2
s))
g gravitation constant (m/s
2
)
h uid enthalpy (kJ/kg)
h
1
cold-side enthalpy (kJ/kg)
h
2
hot-side enthalpy (kJ/kg)
h
2
b
hot-side enthalpy at the maximum mass
ux (kJ/kg)
h
t
height of loop (m)
h
1
restriction loss coefcient at heated
channel inlet
K
2
restriction loss coefcient at heated
channel outlet
p static pressure (N/m
2
)
P period of oscillation (s)
Q total channel power (kW)
Q
b
power when G=G
m
(kW)
Q
s
power on the stability boundary (kW)
z axial distance (m)
Non-dimensional variables
G
*
non-dimensional mass ux
G

m
non-dimensional maximum mass ux
G

s
non-dimensional mass ux on the stabil-
ity boundary
Q

b
non-dimensional bounding power

b
non-dimensional power when
G=0.95G
m
Q

s
non-dimensional power on the stability
boundary
R
*
non-dimensional density
Greek symbols

G
G
m
, a fraction between 0 and 1.
uid density (kg/m
3
)

1
cold-side density upstream of heater
(kg/m
3
)

2
hot-side density (kg/m
3
)

3
uid density downstream of heat sink
(kg/m
3
)

2
b
hot-side density when G=G
m
(kg/m
3
)
inclination to the horizontal (

C
K
2
z
2
C
K
1
z
1
+C
K
3
z
3

_
ght
CK2z2
(m/s)
+

2
+
straight pipe in a once through system and concluded
that supercritical ow instability would be similar to
two-phase ow instability, possessing both the excur-
sive and oscillatory nature. Analytical results were
not given, but formulations were derived to support
the understanding presented. His conclusions on the
parameters that stabilize and destabilize a supercriti-
cal system are, in fact, the same as for a two-phase
system.
Daney et al. (1979) performed an experimental
study of supercritical helium owing in a heated chan-
nel and reported density-wave type of owoscillations.
They purportedly did the rst experimental study of
density-wave oscillations in a cryogenic uid.
Dashkiyev and Rozhalin (1975) studied the stabil-
ity of a systemof parallel steam-generating tubes in the
presence of various types of non-uniformities. Jain et
al. (2003) reported experimental and analytical studies
of CO
2
in a natural-circulation system at supercriti-
cal conditions. No ow instability measurements were
reported, but heat-transfer data were reported.
Chatoorgoon (2001) reported an analytical study of
supercritical water instabilities in a natural-circulation
loop, obtained with the non-linear SPORTS code
(Chatoorgoon, 1986). That study is extended here, and
CO
2
and H
2
are included to determine the effect of dif-
ferent uids. A more recent study by Lomperski et al.
(2004) reported experiments of supercritical CO
2
in a
natural-circulationloop. This is the rst publishedwork
of an experiment at these particular conditions and,
therefore, it is important and relevant. No ow insta-
bilities were obtained in the experiment. The apparent
anomaly between experimental ndings and theoreti-
cal predictions is discussed in Section 8.
2572 V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580
2. Supercritical instabilities
Experimental and analytical studies have been ini-
tiated at the University of Manitoba to further the
understanding of supercritical owinstabilities. Exper-
imental supercritical H
2
O and CO
2
loops are being
constructed, funded by the Canadian Foundation of
Innovation (CFI) and Atomic Energy of Canada Lim-
ited (AECL), which will be used to investigate super-
critical ow stability in vertically oriented, natural and
forced circulation loops.
Chatoorgoon (2001) did an analytical study of
supercritical water ow in a natural-convection loop,
shown in Fig. 1, using the non-linear stability SPORTS
code (Chatoorgoon, 1986). It was found that the steady-
state prole of owrate versus power showed that the
owrate initially increases with power, attains a maxi-
mum and then decreases with further power increases.
It was suggested that the power corresponding to the
maximumsteady-state owrate may be a good approx-
imation of the ow stability boundary. Based on this
Fig. 1. Schematic of loop.
hypothesis, non-dimensional equations were derived
for approximating the stability boundary.
In this study, the SPORTS code is used to conduct
a host of numerical experiments to better understand
supercritical ow instability and the non-dimensional
parameters derived by Chatoorgoon (2001). Different
inlet temperatures, inlet and outlet channel K factors,
different loop heights and heated lengths are exam-
ined. The NIST (2003) property package was used in
SPORTSin lieu of the state equation to close the system
of equations.
3. The SPORTS code
The SPORTS code, which is the principal tool in
this study, was initially developed to study two-phase
ow instability at low pressures and, particularly, the
effect of subcooling boiling on ow instability. The
steep characteristic between void fraction and quality
at lowatmospheric pressures is believedtohave a desta-
bilizing effect on numerical schemes that employ prop-
erty derivatives in its formulation (Chatoorgoon et al.,
1993).
In two-phase applications at near atmospheric con-
ditions, the very small mass and energy content in
the vapor means the uid density is very sensitive to
the energy partitioned between the liquid and vapor.
This increased sensitivity of some property derivatives
across the boiling interface to the partitioned energy
can have a destabilizing effect on a numerical scheme.
The main advantage of the SPORTS formulation is
the avoidance of property derivatives, even though it
employs a relatively simple implicit upwind scheme.
Improvements have been made to the nite-difference
scheme since the original publication (Chatoorgoon,
1986); however, the original procedure used to solve
the governing equations was retained. Property deriva-
tives are avoided in SPORTS formulation by keeping
the state equation separate fromthe conservation equa-
tions and using it to return an improved estimate of the
uid density. The NIST package was implemented into
SPORTS as the state equation and called every time
there was a need for state properties. This approach was
found to produce numerically stable solutions even for
oscillatory ow applications.
In supercritical ow applications, there are also
large changes in some property derivatives across the
V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580 2573
critical point. The large variations in specic heat are
well known, but this difculty is mitigated in SPORTS
by expressing the energy equation in terms of the uid
enthalpy. There is also a large variation in the uid den-
sity across the critical point, but this effect is captured
reasonably well by taking sufciently small section
lengths and ensuring spatial convergence has occurred.
Thus, SPORTS numerical procedure performs well for
computations through the critical and into the super-
critical regions.
4. Non-dimensional stability parameters
A constant area conguration with point heat and
sink sources was considered previously (Chatoorgoon,
2001). This facilitated a closed-form, analytical solu-
tion. The steady-state momentum equation for such a
system is:

z
_
G
2

_
=
p
z
+g sin C
K
_
G
2

_
(1)
where
C
K
z K +
fz
2D
(2)
K is the restriction loss coefcient and f is the friction
factor. In the publication by Chatoorgoon (2001), the
effect of K was wrapped up in the f term. In this paper
the more conventional approach is followed and they
are kept distinct and separate.
The mass ux, G, is given by (Chatoorgoon, 2001):
G
2
=
gh
t
(
1

2
)
C
K
1
z
1
+C
K
3
z
3

1
+
C
K
2
z
2

2
(3)
By postulating that the stability boundary power,
Q
b
, is the power corresponding to the maximum
owrate, setting
G
Q
= 0 and G=G
m
yields:
_
G
m

2
b
_
2
=
gh
t
C
K
2
z
2
=
2
(4)
G
m
=
2
b
(5)

gh
t
C
K
2
z
2
(6)
The equation of state on the supercritical tempera-
ture side was assumed to be of the form (Chatoorgoon,
2001):

2
=
B
h

2
2
(7)
For water, B=1.7418 10
23
and
2
=3.277 when h
2
is in J/kg.
Thus, it was obtained:
Q

b
=
Q
b
A
1
h
1

(8)
Q

b
=
_
B

_
1/
2
1, (9)
where
B

=
B

1
h

2
1
(10)
= +
_

2
+ (11)
=
C
K
2
z
2
C
K
1
z
1
+C
K
3
z
3
(12)
The normalized peak steady-state ow rate is:
G

m
= (13)
where
G

m

G
m

(14)
R

b
= (15)
where
R

b


2
b

1
(16)
The aim of this study is to assess the validity of the
non-dimensional parameters Q

b
, G

m
and R

b
(Eqs. (8),
(13) and (15)) on the stability boundary by numerically
simulating a large number of cases covering a wide
range of conditions relevant to a nuclear reactor.
5. The geometry
The geometry chosen is similar to the one reported
by Chatoorgoon (2001). It is a simple, constant area,
single channel ow loop oriented vertically, as shown
in Fig. 1. The ow is driven by natural convection.
2574 V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580
Heat is applied on the central portion of the lower
horizontal segment and removed from the central por-
tion of the upper horizontal segment. In all cases,
the amount of heat removed was equal to the amount
applied so there was no net energy gain to the system
over time. Also, energy was assumed to be deposited
directly to the ow and, therefore, the heat-transfer
between the wall and the ow was not modeled. In
future studies, the heat capacity of the pipes will be
included to ascertain how much it impacts on the sta-
bilitycharacteristics. The heat removal lengthwas xed
at 1 m for all cases.
Two heated lengths of 1 m and 2 m were used. Even
though the non-dimensional parameters being tested
here were derived originally (Chatoorgoon, 2001) for
a point source and sink, nite heated lengths are con-
sidered here for greater realism.
The following parameters were varied: inlet tem-
perature, channel inlet and outlet K factors, vertical
loop height, channel heated length and the applied
power. Friction factors were calculated using the stan-
dardformulae for laminar andturbulent subcritical ow
embeddedinthe code, as the most appropriate formulae
for supercritical ow is still a matter of debate. While
this may result in small errors in the pipe pressure drop
when the ow is supercritical, this is expected to be of
little consequence since most of the pressure drop in
these simulations were dominated by the K factors. A
different friction formula would yield a very slightly
different steady-state ow rate for a given power, but
the effect on the non-dimensional parameters studied
here, and the ensuing conclusions, are not expected to
be affected.
6. Cases studied
The cases studied numerically are given in Table 1.
Atotal of 94 cases were simulated. These include three
inlet temperatures of 340

C, 350

Cand 360

C; chan-
nel inlet and outlet K factors varying from 0.5 to 10.0;
three vertical loop heights of 10 m, 14 mand 17 m; two
heated lengths of 1 mand 2 m. The systempressure was
25 MPa and the critical temperature was 374.3

C.
7. Findings
The validity of the non-dimensional parameters,Q

b
,
G

m
and R

b
, on the stability boundary is now assessed
by determining howwell the predicted stability bound-
ary conditions correlate with the non-dimensional
parameters.
7.1. Boundary conditions
The system stability would be, naturally, very
dependent on the boundary conditions used. In these
simulations, the inlet temperature, inlet and outlet pres-
sures were assumed constant.
Table 1
Cases simulated
L
TS
(m) T
IN
(

C) h
t
(m) K
1
K
2
No. of runs
1 350 10 1.010.0 1.0 10
1 350 14 1.010.0 0.5 10
1 350 17 0.53.0 0.5 4
1 340 14 1.010.0 1.0 10
1 340 14 5.0 8.0 1
1 360 14 1.06.0 1.0 6
1 360 14 8.09.0 10.0 2
2 350 10 1.010.0 0.5 10
2 350 14 1.010.0 1.0 10
2 350 17 1.010.0 1.0 10
2 340 14 1.04.0 5.0 4
2 340 14 1.05.0 4.0 5
2 360 14 1.010.0 1.0 10
2 360 14 8.09.0 8.0 2
V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580 2575
7.2. Stability boundary
The SPORTS code determines system stability
through a non-linear, time transient solution of the
governing equations. At time zero, a perturbation is
introduced into the inlet ow velocity and a system
transient is executed. From this time transient, the sys-
temstability is determined (Chatoorgoon, 1986, 2001).
Fig. 2 gives the result of the rst case in Table 1. It is
clear that at Q
s
=5.0 MW the system is slightly stable,
at 5.3 MWthe systemis just about neutrally stable, and
at 5.4 MW the system is slightly unstable, revealed by
Fig. 2. Transient case showing the stability boundary close to
5.3 MW.
Fig. 3. G

s
vs. G

m
for 1 m heated length.
the onset of diverging oscillations. Thus, 5.3 MW was
taken to be the stability boundary. At higher powers,
the oscillations diverge more rapidly (not shown).
7.3. Results
7.3.1. Non-dimensional parameters
Figs. 3 and 4 are plots of G

s
versus G

m
, where
G

s
_

G
s

_
is the non-dimensional value of G
s
the
value of mass ux on the stability boundary as cal-
culated from a SPORTS stability analysis. Fig. 3 is
for 1 m heated length and Fig. 4 is for 2 m heated
length. A total of 94 cases are plotted for the different
Fig. 4. G

s
vs. G

m
for 2 m heated length.
2576 V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580
Fig. 5. G

s
vs. for 1 m heated length.
channel inlet temperatures, channel inlet and outlet K
factors, and loop heights. It is evident that the agree-
ment between G

s
and G

m
is very good. Noting that
G

m
is the non-dimensional peak steady-state mass ux
as determined from a steady-state analysis, it is obvi-
ous that G

m
is a very good approximation of G

s
the
non-dimensional mass ux at the stability boundary.
The lower dotted line in both gures is the line cor-
responding to 95% of G

m
. This means that the lines
G

s
= G

m
and G

s
= 0.95G

m
bracket all the stability
predictions. Therefore, the stability boundary ow rate
can be predicted to within 95% accuracy by using the
peak steady-state ow rate as the stability boundary
ow rate.
Figs. 5 and 6 are plots of G

s
versus , where is
given by Eqs. (11) and (12). The parameter is actu-
ally calculated by SPORTS, which returns C
K
for every
node. It is a combination of the K factor and the fric-
tional pipe loss factor. z
1
is taken to be the length from
inlet to the critical point in the heater section, z
1
is taken
Fig. 6. G

s
vs. for 2 m heated length.
Fig. 7. R

s
vs. for 1 m heated length.
to be the length fromthe critical point in the heater sec-
tion to the critical point in the heat-exchanger, while z
3
is taken to be the length from the critical point in the
heat-exchanger to the outlet.
Fig. 5 is for 1 m heated length and Fig. 6 is for
2 m heated length. This data also correlates very well.
Noting that is mainly a geometric parameter, it is
interesting that G

s
is dened solely by this parameter.
Figs. 7 and 8 are plots of R

s
_


2
s

1
_
versus
for 1 m and 2 m heated lengths, respectively.
2
is the
hot side density when the system is at the stability
boundary. It is evident that the non-dimensional hot-
side density at the stability boundary correlates well
with , as predicted by Eq. (15). As before, the dotted
lines correspond to setting G

s
= 0.95G

m
. It is encour-
aging to see that all the numerical data fall within the
dotted lines.
Fig. 8. R

s
vs. for 2 m heated length.
V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580 2577
Fig. 9. Q

s
vs. {(B

/)
1/
2
1} for 1 m heated length.
Figs. 9 and 10 are plots of Q

s
versus
__
B

_
1/
2
1
_
, where Q

s
_

Q
s
A
1
h
1

_
is the
non-dimensional value of Q
s
the actual stability
boundary power as obtained by a stability analysis.
Even though the data is centered on the theoretical
line, there is a large scatter in Q

s
. The reason for the
large scatter in Q

s
is the almost insensitive variation
of ow rate with power near the peak of the ow rate
versus power steady-state prole (see Fig. 11). This
is due to the atness of the prole near the peak ow
rate.
The dotted lines in Figs. 9 and 10 plot

Q

b
, where

b
is derived in Part II of this study. It is the non-
dimensional value of Q corresponding to G=0.95G
m
.
Fig. 10. Q

s
vs. {(B

/)
1/
2
1} for 2 m heated length.
It is given by:

b
=
__
B

_
1/
2
1
_
(17)
where

=
2
+
_
1
2
2
_

1
2
_
(1
2
){(1
2
) +4
2
(1 )} (18)
and
=
G
G
m
(19)
When =1, G=G
m
and

= . A value of =0.95
was used to derive the dotted lines in the gures.
The rationale for

Q

b
is explained in Part II
of this study. Because all the stability data fall
between the dotted lines dened by

Q

b
, this means
that Q

s
is bracketed between the two steady-state
powers corresponding to G=0.95G
m
. This nding
suggests that the stability boundary power can lie
anywhere in the stability boundary region denoted
in Fig. 11 the edges of the region being dened
by the pair of steady-state powers corresponding to
G=0.95G
m
.
Figs. 12 and 13 are plots of Q

s
versus Q

b
. Fig. 12 is
for a 1 m heated length and Fig. 13 is for a 2 m heated
Fig. 11. Steady-state prole showing stable, unstable and stability
boundary regions.
2578 V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580
Fig. 12. Q

s
vs. Q

b
for 1 m heated length.
length. Q

b
is the non-dimensional power obtained from
a steady-state analysis and is obtained from the power
corresponding to the peak steady-state owrate. While
Q

b
is obtained from a steady-state analysis, Q

s
is
obtained from a stability analysis. As in Figs. 9 and 10,
the data is scattered around the =1 line and is brack-
eted by the =0.95 dotted lines. Thus, the data is
reasonably well correlated. Otherwise, the explana-
tions for this case are similar to the explanations given
above for Figs. 9 and 10.
7.3.2. Period of oscillation
The period of oscillation, P, is given in the last col-
umn in Tables 2 and 3. It varies from 1.7 to 8.2 s. A
discussion on the oscillation period is given in Part II.
It is omitted here for the sake of brevity and because it
is not important to the validation exercise of the chosen
non-dimensional parameters.
Fig. 13. Q

s
vs. Q

b
for 2 m heated length.
Table 2
1 m Heated length cases
T
IN
(

C) h
t
(m) K
1
K
2
Q
s
(MW) G
s
(kg/
(m
2
s))
P (s)
340 14 1 1 5.30 2215.86 7.5
2 1 5.50 2211.13 6.5
3 1 6.10 2208.31 5.7
4 1 7.40 2175.48 2.7
5 1 7.43 2108.21 3.2
6 1 7.45 2047.59 2.4
7 1 7.48 1992.77 2.2
8 1 7.50 1942.76 2.1
9 1 7.55 1896.92 2.0
10 1 7.60 1853.91 2.0
5 8 4.70 1419.14 3.2
350 10 1 1 5.50 2090.12 4.5
2 1 5.80 2002.83 3.5
3 1 6.60 1922.74 2.5
4 1 7.30 1832.50 3.0
5 1 7.40 1764.90 2.0
6 1 7.50 1740.20 1.9
7 1 7.60 1648.96 1.8
8 1 7.80 1594.05 1.8
9 1 8.00 1542.53 1.7
10 1 8.30 1490.54 1.7
14 1 0.5 4.80 2311.18 7.0
2 0.5 5.10 2244.74 6.5
3 0.5 5.40 2184.27 7.0
4 0.5 5.80 2134.29 6.8
5 0.5 6.20 2083.06 6.0
6 0.5 7.10 2041.15 2.0
7 0.5 7.30 1988.09 1.9
8 0.5 7.40 1929.42 1.8
9 0.5 7.45 1880.57 1.8
10 0.5 7.50 1835.82 1.8
17 0.5 0.5 5.50 2523.95 7.0
1 0.5 5.70 2487.97 6.8
2 0.5 6.20 2425.91 6.2
3 0.5 6.70 2358.30 5.5
360 14 1 1 6.10 2187.18 5.0
2 1 6.80 2091.40 4.0
3 1 8.00 1965.79 2.0
4 1 8.10 1892.92 2.0
5 1 8.20 1827.66 1.8
6 1 8.30 1866.05 1.7
8 10 4.30 1109.03 2.2
9 10 4.40 1087.79 2.2
7.3.3. Spatial and temporal convergence
For the results presented here, the geometry was
divided into 10 cm section lengths with a time step of
about 1/20 of the period of oscillation. Time steps less
than one-fteenth of the oscillation period was found
V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580 2579
Table 3
2 m Heated length cases
T
IN
(

C) h
t
(m) K
1
K
2
Q
s
(MW) G
s
(kg/
(m
2
s))
P (s)
340 14 1 4 4.20 1860.55 7.9
2 4 4.30 1806.13 7.7
3 4 4.40 1762.78 7.7
4 4 4.50 1723.43 8.0
1 5 4.20 1694.53 6.0
2 5 4.40 1658.30 6.2
3 5 4.80 1621.28 6.3
4 5 4.90 1587.09 6.2
350 10 1 0.5 4.50 2126.13 6.5
2 0.5 4.65 2037.78 6.1
3 0.5 4.70 1954.79 6.0
4 0.5 5.10 1908.20 5.5
5 0.5 5.25 1849.33 5.5
6 0.5 5.35 1794.11 5.3
7 0.5 6.00 1759.91 5.4
8 0.5 6.10 1711.43 4.0
9 0.5 6.40 1668.24 4.0
10 0.5 6.80 1625.33 4.2
14 1 1 5.10 2249.01 6.8
2 1 5.50 2188.24 6.2
3 1 5.80 2124.98 6.1
4 1 6.10 2065.55 6.1
5 1 6.50 2010.00 4.5
6 1 6.80 1954.35 4.2
7 1 7.35 1896.23 3.6
8 1 7.60 1841.42 3.3
9 1 8.20 1780.53 3.2
10 1 8.30 1727.32 3.0
17 1 1 4.60 2285.73 8.2
2 1 4.80 2247.83 7.4
3 1 5.00 2175.18 8.0
4 1 5.50 2148.38 6.8
5 1 5.70 2100.16 6.4
6 1 5.90 2054.94 6.1
7 1 6.30 2022.40 5.2
8 1 6.60 1976.78 5.2
9 1 6.75 1937.01 5.2
10 1 7.50 1892.22 5.3
360 14 1 1 4.60 2130.23 6.8
2 1 5.00 2069.01 6.2
3 1 5.20 2002.59 6.0
4 1 5.40 1942.64 5.6
5 1 5.50 1889.48 5.0
6 1 5.90 1837.43 4.7
7 1 6.20 1789.93 4.5
8 1 6.40 1742.58 4.5
9 1 6.80 1694.80 4.5
10 1 7.00 1651.93 4.0
8 8 5.10 1139.33 2.8
9 8 5.20 1123.78 2.9
from past experience to yield temporal convergence of
the ensuing transient. To ensure spatial convergence
was attained, some of the runs were repeated with
5 cm section lengths instead of the usual 10 cm. This
yielded very similar predictions of the stability bound-
ary, thereby establishing that spatial convergence was,
in fact, attained.
8. Temperature versus pressure boundary
condition
The experimental ndings of Lomperski et al.
(2004) are sufciently important to warrant this sep-
arate commentary. In that study, a natural circulation
experiment with supercritical CO
2
was conducted, but
noowinstabilities were foundat powers near the peak,
and even on the negative slope, of the owratepower
prole. Thus, the followingthoughts are sharedfor con-
sideration:
(a) Chatoorgoon and Upadhye (2005) developed a lin-
ear stability model that attained at least a 95%
agreement with SPORTS predictions in all sta-
bility cases examined. Since the linear model is
exempt from the numerical problems associated
with nite-difference schemes, and is considered
proven methodology, it can be concluded that the
instability predictions given here are real and not
a result of any numerical procedure, as was sug-
gested by Lomperski et al. (2004).
(b) The nding here is that the stability boundary
power, Q
s
, does not always coincide with the
power, Q
b
, at the peak steady-state owrate. In
fact, sometimes Q
s
is greater than Q
b
by 30%
and sometimes Q
s
is less than Q
b
, as is evident
from Figs. 12 and 13. This means that some points
on the negative slope in Fig. 11 can be stable, as
well as some points on the positive slope in Fig. 11
can be unstable.
(c) Amore likely explanation for the anomaly between
Lomperski et al. (2004) experimental results and
these predictions is the different boundary condi-
tion used in the code analyses and in the exper-
iment. In these analyses, a constant loop outlet
pressure boundary condition was assumed and the
outlet ow temperature from the heat exchanger
was allowed to vary. In the ANL experiments,
2580 V. Chatoorgoon et al. / Nuclear Engineering and Design 235 (2005) 25702580
a constant heat-exchanger outlet temperature was
imposed and, hence, did not vary. Aconstant outlet
temperature boundary condition would yield dif-
ferent stability predictions to a constant outlet pres-
sure boundary condition. In fact, for a constant inlet
and constant outlet temperature boundary condi-
tion, the momentumequation is not needed to solve
for the stability of the system and the system is, in
fact, quite stable. This was shown to be the case by
Chatoorgoon and Upadhye (2005).
9. Conclusions
A large number of numerical studies were per-
formed to better understand the non-dimensional
parameters dening the ow stability boundary of
supercritical water innatural-convectionloops. Arange
of inlet and outlet K-factors, various inlet temperatures,
heated lengths and vertical loop heights were used in
the study. The following were found:
(1) G

m
is a very good approximation of G

s
. The accu-
racy is about 95%.
(2) G

s
is well approximated by the geometric param-
eter , with an accuracy of about 95%.
(3) R

s
is also well approximated by , with an accu-
racy of about 95%.
(4) Q

s
can be approximated by Q

b
(see also point
(5) below), which in turn can be approximated by
__
B

_
1/
2
1
_
.
(5) There is a larger scatter between Q

s
and Q

b
than
between the other foregoing parameters. However,
all the Q

s
data are encompassed within the pair of

b
values Eq. (17) dened as the non-dimensional
power values corresponding to 95% of G

m
.
(6) The non-dimensional parameters G

m
, R

b
and Q

b
,
which were originally derived for a point heat
source and sink situation, appear to be valid param-
eters dening the stability boundary in supercrit-
ical natural-convection water ow system with
distributed heating and cooling.
References
Chatoorgoon, V., 1986. A simple thermalhydraulic stability code.
Nucl. Eng. Des. 93, 5167.
Chatoorgoon, V., Dimmick, G.R., Carver, M.B., 1993. Modelling of
low pressure subcooled boiling instability experiments, instabil-
ity in two-phase systems. ASME Winter Meeting.
Chatoorgoon, V., 2001. Stability of supercritical uid owin a single-
channel natural-convection loop. Int. J. Heat Mass Transfer 44,
19631972.
Chatoorgoon, V., Upadhye, P., 2005. Analytical studies of supercrit-
ical ow instability in natural convection loops. In: 11th Interna-
tional Topical Meeting on Nuclear Reactor Thermal-Hydraulics
(NURETH-11), Avignon, France, October 26.
Cornelius, A.J., Parker, J.D., 1965. Heat transfer instabilities near
the critical point. In: Proceedings of the 1965 Heat Transfer and
Fluid Mechanics Institute. Stanford University Press, pp. 317
329.
Daney, D.E., Ludtke, P.R., Jones, M.C., 1979. An experimental study
of thermally-induced ow oscillations in supercritical helium.
ASME J. Heat Transfer 101 (February), 914.
Dashkiyev, Y.G., Rozhalin, V.P., 1975. Thermo-hydraulic stability of
a system of steam-generating channels with super-critical pres-
sure. Heat Transfer Sov. Res. 7 (SeptemberOctober (5)).
Dimmick, G.D., Chatoorgoon, V., Khartabil, H.F., Duffey, R.B.,
2002. Natural-convection studies for advanced CANDU reactor
concepts. Nucl. Eng. Des. 215, 2738.
Harden D.G., Boggs, J.H., 1964. Transient ow characteristics of a
natural-circulation loop operated in the critical region. In: Pro-
ceedings of the 1964Heat Transfer andFluidMechanics Institute.
Stanford University Press, pp. 3850.
Jain, R., et al., 2003. Studies of natural circulation heat transfer and
ow stability of a supercritical uid. In: 10th International Topi-
cal Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-
10), Seoul, Korea, October 59.
Lomperski, S., Cho, D., Jain, R., Corradini, M.L., 2004. Stability of
a natural circulation loop with a uid heated through the ther-
modynamic pseudocritical point. In: Proceedings of ICAPP04,
Pittsburgh, PA, USA, June 1317, Paper 4268.
NIST/ASMESteamProperties, 2003. Version 7.0. US Dept. of Com-
merce, Technology Administration, National Institute of Stan-
dards and Technology.
Walker, B.J., Harden, D.G., 1964. The density effect model: pre-
diction and verication of the ow oscillation threshold in a
natural-circulation loop operating near the critical point. ASME
Paper No. 64-WA/HT-23.
Zuber, N., 1966. An analysis of thermally induced owoscillations in
the near-critical and super-critical thermodynamic region. Rept.
NASA-CR-80609. Research and Development Center, General
Electric Company, Schenectady, New York, USA, May 25.

Potrebbero piacerti anche