Você está na página 1de 14

Pergamon

Chemical Engineerin 0 Science, Vol. 51, No. 2, pp. 295-308, 1996


Copyright 1995 Elsevier Science Ltd
Printed in Great Britain. All fights reserved
0009 2509/96 $9.50 + 0.00

0009-2509(95)00256-1

A S I M P L I F I E D M O D E L FOR REAL GAS E X P A N S I O N


B E T W E E N TWO RESERVOIRS C O N N E C T E D BY A THIN TUBE
S. CHARTON ~, V. BLET~ and J. P. CORRIOU t*
~LSGC-CNRS-ENSIC, BP451, 54001 Nancy Cedex, France; ~CEA, BP12, 91680 Bruy~res-le-Chfitel,
France
(First received 27 December 1994; accepted in revised form 19 July 1995)
Abstract--A simplified model is derived for simulating gas blowdown through a thin and long tube
connecting a high-pressure gas-filledreservoir to a vacuum vessel. During the depressurization process, the
flow is assumed to be quasi-steady and approximated as one-dimensional.The transient compressibleflow
can therefore be solved analytically. The model accounts for gas nonideality and heat transfer with the
ambient in both reservoirs, and for wall friction in the pipe. An experimental study has been carried out
using helium and deuterium in order to validate this simplifiedapproach. The discharge of the pressure
vessel being quite rapid (a few seconds long), the measurements reliability and the required adjustments are
discussed. The model predictions are shown to be in good agreement with our experimental results.

INTRODUCTION
The depressurization of large pressure vessels occurs
in process plants either accidentally or voluntarily.
Consequently, and for the sake of security, the system
behavior must be understood. However, the related
literature is not abundant.
Concerning the vessels themselves, pressure-relief
devices must be sized, with respect to the highest
pressure admitted in the vessel, in order to ensure
a sufficient exiting flow rate in case, for instance, of
runaway reactions. Furthermore, blowdown time
should be evaluated through a given orifice (Leung,
1986). Internal phenomena are also to be described,
and particularly the thermal history: the wall temperature may indeed reach the ductile-brittle transition
point of the metal. Assuming that the flow is steady
and using a constant heat transfer coefficient between
the vessel wall and the fluid, Xia et al. (1993) derived
a simplified model to predict the temperature and the
pressure evolutions during the discharge. A more
complete computer model has been developed by
Haque et al. (1992), which simulates the tank pressure
and temperatures histories, as well as the transient
flow rate, during a rapid depressurization process.
On the other hand, Levenspiel (1977) studied the
atmospheric discharge of an infinite reservoir through a
pipe. Focusing on the phenomena induced by the pipe,
he determined, by means of thermodynamic considerations, the adiabatic, frictional and steady flow rate of a
perfect gas for various tube size and reservoir pressures.
This paper is related to the simultaneous occurrence of these two extreme situations: the transfer of
a real gas from a high-pressure vessel to a low-pressure one through a thin tube. It is then important to
describe correctly the discharge of the container, as

*Corresponding author.

well as the simultaneous filling of the receiver, in terms


of mass and temperature variations, since they both
define the boundary conditions of the transient flow
occurring in the pipe.
A rigorous investigation, concerned with dynamic
flow-patterns and local temperature fields, would imply the resolution of the mass, energy and momentum
balances of the compressible fluid in the whole system.
Many schemes are available for compressible pipe
flow calculations (Fletcher, 1991) although not ensuring the correctness of the calculated results because of
the system complexity. Most of these algorithms are
suitable when inflow and outflow boundary conditions are specified. However, these latter are often
difficult to define and the computational task would
become tricky and time consuming in our configuration. Indeed, because of the abrupt constriction at the
tube inlet, the use of a fine grid would be essential, at
least locally. Furthermore, another difficulty arises
due to the significant pressure gradient prevailing
between the two enclosures: pressure waves are generated, which propagate in both directions of the tube,
and reflect on the vessels' walls until a steady flow
regime is achieved. The numerical difficulties encountered when handling such discontinuities can be overcome by means of the characteristics theory. Issa and
Spalding (1972) derived a hybrid scheme which combines the use of a rectangular grid with the use of
characteristics. It is therefore particulary efficient for
shock-tube simulations.
The object of this paper is to present a simplified
way of modelling this specific transfer, while keeping
a sufficient accuracy for chemical engineering purposes.
In order to reach an adequate compromise between
calculation accuracy and computation efficiency, the
choice of the model assumptions is of prime importance, in order to account only for the most significant
effects and to neglect secondary phenomena.

295

296

S. CHARTON e t

The derived model is inspired from the analytical


procedure described by Levenspiel (1977) insofar as
a steady, adiabatic and frictional flow of a perfect gas
is considered at each time step in the tube. Contrary
to this previously studied case, the flow system is such
that temporal variations of pressure and temperature
take place in both the discharging and downstream
reservoirs. Hence, rigorous transient mass and energy
balances, accounting for real gas behavior and heat
transfer with the ambient, are implemented in the
model for the reservoirs treatment. The Soave cubic
equation of state is used to estimate compressibility
factors and departure functions.
An experimental study of the process has been
carried out in order to test the model reliability. Two
gases of close mole weight were chosen: helium and
deuterium, respectively monoatomic and diatomic.
Initial pressures ranging from 10 to 45 MPa were
investigated for both a long (1.1 m) and a short (0.3 m)
tube. As fast variations are involved in the studied
process, compared to the dynamics of the sensors,
measurement distortions are encountered. For this
reason, prior to any confrontation of the model predictions with experimental data, the raw signals have
been treated to account for the sensors characteristics.
This experimental investigation validates the transfer-time prediction and reveals, in each case, a good
agreement between predicted and measured profiles
in the vessels.
2. T H E O R E T I C A L D E V E L O P M E N T

The model concerns the transient transfer of a real


gas from a high-pressure reservoir to a low-pressure
reservoir through a thin and long tube (Fig. 1). Assumptions made in the development of the model are
listed in the following:
1. In the vessels:

(a) For the purpose of the simplified approach,


temperature and pressure are assumed to be spatially
uniform. So, homogeneous conditions are prevailing
in both vessels.
(b) Moreover, the vessels are large with respect to
the tube section. Consequently, the gas velocity is
neglected and stagnant conditions are admitted.
(c) The treatment of heat losses is simplified by
considering a constant wall temperature, equal to the

al.

temperature of the surroundings. Indeed, the transfer


rapidity prevents the heavy metallic-shell temperature
from being significantly affected by the thermal variations, however important they are, of the small
amount of gas involved, in accordance with the total
heat-balance: (mCe) . . . . ImTmetal = (mCe)gasATgas.
(d) Like in industrial vessels discharge, we assume
that natural convection prevails in the pressure container (Haque et al., 1992), since it is driven by the
density gradients generated by the depressurization.
(e) In the downstream reservoir, to be consistent
with the stagnant conditions specified, we will also
assume, according to the Boussinesq approximation,
that the gas motion results only from the density gradients, and hence that natural convection is dominant.
These two last assumptions will be discussed later.
(f) The global heat transfer coefficient is calculated
in both vessels by eq. (1), valid for isothermal surfaces
(Cess, 1973):

-N'-uU = a l R a

Nu

'25

aERa '33

for Ra <~ 109


for Ra > 109

where the global Nusselt number and the Rayleigh


number are respectively defined by
S~ =

QttalfluxL
2S(Tw-

T~)

__

h~L
2

and
R a = Gr Pr =

gfl(Tw - T ~ ) p 2 L 3 Cplt
It 2

L is the characteristic length of the exchange surface,


Tw is the wall temperature, T~ the bulk gas temperature, al is set to 0.47 for a cylinder of diameter L and
0.49 for a sphere, and a2 = 0.1 for a cylinder of length
L and for a sphere.
The dependence of relation (1) on the Rayleigh
number allows to account for the onset of turbulent
convection (still generated by the density gradients) in
the reservoirs.
Since the gas properties depend on temperature, the
reference temperature defined by Sparrow and Gregg
(Cess, 1973) is adopted to estimate C v, #, 2, and p:
T = T w - 0.38(Tw - T).

(2)

Only the dilatation factor fl = (1/p)(dp/aT)v is calculated at the bulk temperature.

RECEIVER

i
I

x]

'~11
Hish-Pretsu~

(1)

gl
Vacuum Conditiom

Fig. 1. Schematic illustration of the process.

A simplified model for real gas expansion

297

1.40

"O 1.30

~J

1.20

!
t..

.=
1.10

1.00
L

E 0.90
0.90

1.00

1.10

1.20

1.30

1.40

Compressibility Factor - Experimental Values

Fig. 2. Comparison of compressibility factors (Z = PV/RT) in the range 102 ~<P (Pa) ~<3.5 x 107 and
273.15 ~< T (K) ~<373.15 for helium, and 102 ~<P (Pa) ~<3.5 x 107 and 223.15 ~< T (K) ~<373.15 for
deuterium. One point materializes one couple {pressure, temperature}.
(g) Finally, the real gas behavior is assumed to be
correctly described by the Soave equation of state.
This cubic equation is particularly convenient for
computational purposes. The derivation of the departure functions from ideal gas behavior is described in
Reid (1987).
Comparison of estimated compressibility factors
with available experimental values (Fig. 2) points out
that the use of this EOS is consistent with an accuracy
better than 5% for helium and 3% for deuterium. We
also notice that in the range of conditions investigated, the departure from perfect gas behavior (Z = 1)
in the reservoirs can rise up to 25%.

2. In the tube:
(a) The flow is quasi-steady. Indeed, as it was mentioned previously, and will be illustrated later, at the
beginning of the transfer, a rarefaction wave proceeds
upstream in the tube, while a compression or shock
wave proceeds downstream, thus ensuring that pressure conditions consistent with the frictional flow are
prevailing between the tube openings. Under such
choked conditions, the transient flow turns to a quasisteady one.
Thanks to this major assumption, the numerical
task is greatly simplified.
(b) Unlike the assumption taken inside the vessels,
the ideal gas law is assumed, thus allowing the derivation of an analytical procedure.
(c) Because of the small tube section, radial variations of the fluid properties are negligible compared
to the longitudinal ones. Therefore, the flow is treated
as one-dimensional.
(d) Finally, the flow is supposed to be adiabatic.
This is reasonable, first of all, because the gas resi-

dence time in the tube is very short (around 3 ms).


Furthermore, at the tube wall, the gas velocity is
almost zero; its temperature at the wall is therefore
equal to the stagnation temperature T o of the flowing
gas. Hence, as mentioned by Levenspiel (1977), the
constant wall temperature in the tube is better represented by the adiabatic flow assumption: d T /dz = O.
(e) To set the inlet conditions, we assume that the
fluid expands isentropically at the pressure-vessel oririce. The significant effect is indeed the mechanical
work of the tube itself and not the pressure drop
induced by the orifice.
Based upon these assumptions, mass and energy
balances in the reservoirs yield:

In the pressure vessel:

dpi

V1 - -

dt

= - p2u2[2

dU~=dt - p 2 u 2 ( h 2 + ~ )

(3)

+hwSw(Tw- T1)" (4)

In the downstream reservoir:


do.
V4 - ~ = p3u3~
d t = psU3 h3 +

+ hwSw(rw- r4)

(5)
(6)

where U refers to the internal energy of the gas contained in the vessel and h2, h3 to the specific enthalpy
of the gas respectively at the tube inlet and outlet.
They are estimated from the corresponding ideal gas

298

S. CHARTON et al.

properties by means of the departure functions. Pressure, temperature and density are related by the real
gas law: P = p Z ( R / M ) T where Z is given by the cubic
equation of Soave (1980):
Z 3 - - Z 2 "~ [A* -- B*(1 + B*)]Z - A ' B * = 0

extremities and leads to


+ 1 In ( J/2{1 + [(7 - 1)/2] J/2}'~
2

\M/32 {1 + [(y

(7)
-

with A* = aP/R2T 2 and B* = bP/RT.


In the tube, the flow is described by the following
mass, momentum and energy conservation equations
d(pu)
= 0
dx

(8)

d(pu u) dP 2Cf
2
dx + -~x + ---ff- pu = 0
d--x pu h + ~ -

P3
P2

---if-

(15)

J[2 N/~ + [(Y -- 1)/2] d/2


J43
+ [(y 1)/2] j [ 2 -

(16)

The inlet boundary conditions have been previously defined by assuming an isentropic expansion
of the gas leaving the container, that is

(9)

=0

1)/2] J/~}J

, P1 _ P2
~7-=7
Pl
P2

(10)

ht = 112 -{ us
2

and the ideal gas law


R

P = p - ~ T.

(11)

In eq. (9), C: is the friction factor. It is commonly


admitted that incompressible fluid dynamic correlations remain valid for subsonic flows of compressible
gas (Shapiro, 1953). Churchill's correlation has been
chosen for its accuracy in both laminar and turbulent
flow regimes:
C.r V ( 8 " ~ ~2
1
1 ~/2
-2- = [_\-~eJ + (A + B) 3/2

(12)

with

and
(37,530~ 10
B = \----RT--e) "
Re = puD/# is the flow Reynolds number, e is the pipe
roughness.
Introducing the dimensionless velocity
,A

For an ideal gas, enthalpy is only temperature dependent: dh = CvdT. It comes in terms of the Mach
number:
PI=P2

T , = T2(1 + ~ _ ~ ~ 2 ) .

~h

hO=hc+IC-2
2i
U [

1 - d 2
d~1
~g2{ 1 + [ ( T - 1)/2] ~//a} #//
1 +(7+1).//2
-

(18)

Finally, eqs (15)-(18) and eq. (11) constitute a set of


five equations involving six unknowns, Pa, Pa,
T2, T3, ,g/2 and ~3, where subscripts refer to points
indicated in Fig. 1. All but one can be deduced from
P1 and Tt. Indeed, the determination of,//2 requires
another set of boundary conditions since ~/2 depends
both upon the upstream and downstream flow
boundaries.
The one-dimensional, adiabatic and frictional flow
we are dealing with is known as Fanno flow (Shapiro,
1953). It is characterized by a constant stagnation
enthalpy h = h + u=/2, and is usually illustrated in
an enthalpy/entropy diagram by the set of constant
flow rate curves. The upper branch of each Fanno
curve refers to subsonic flows, while the lower one is
only described by supersonic flows (Fig. 3).

eqs (8)-(11) yield

(17)

dP

7 - 1 j / 2 ) ~/(~-1)
1 + ----~-

2yCrdx
D

"~"

(13)

de//

1 + [(~ - 1 ) / 2 ] ~ '~ ~

(14)

"

Equations (13) and (14) describe respectively the


Mach number variations along the tube for the
adiabatic and frictional flow, and the corresponding
pressure variations with respect to ./. Their analytical integration can be derived between the two

C*

~s

Fig. 3. Fanno curves in the enthalpy/entropy diagram.

A simplified model for real gas expansion


Since there are no divergent devices in the flow
system of interest here, we are only concerned with
subsonic flows: ~ ~< 1.
Let F s represent the steady flow at the pipe inlet
2 at time t. As the gas progresses in the tube, its
velocity is enhanced by frictional pressure drop and
F S moves to the right on the upper branch of the
Fanno line. If the tube is long enough, the sonic
velocity associated to the critical point C* is reached:
F S = 3 = C*, but cannot be passed beyond. Indeed,
this transition from a subsonic to a supersonic flow
regime, together with an entropy diminution, is
actually a violation of the second law of thermodynamics. Instead, choking occurs and acts to decrease the flow rate until a steady-state solution again
becomes possible with 3 = C*.
From these thermodynamic considerations, it
comes out that, regardless of the pressure ratio P1/P4,
the highest velocity attainable is restricted to the
speed of sound c = ~ ) T ,
and can only be
reached at the end of the pipe. Thus, under choking
conditions, the exhaust pressure is higher than the
backpressure (the receiver pressure in this case).
The missing set of boundary conditions is deduced
from the foregoing discussion. Substituting eq. (17) in
eq. (16) together with ~3 = 1 yields the gas pressure
at the critical point C*:
p*

/ 2 ~ g 2 { 1 + [(~' - 1)/2],/(2}~1+y)/.-~'~

~-1 = ~ /

y+l

(19)

Consequently, under choking conditions, the flow is


such that
P1

>(P3 =P*)>IP4

JCa = 1

Container

7
X

299

otherwise
P1 > (P3 = P4)/> P*
J//3 < 1.

Using these relations, the system of equations (15),


(16) and (17) is solved iteratively in order to find J/2 at
each time step. The other flow properties are then
easily derived.
3. EXPERIMENTAL

3.1. Equipment
The experimental apparatus (entirely made of stainless steel) is constituted by a set of two spherical
instrumented reservoirs, connected by a thin tube
(Fig. 4). One of them is filled at a relatively high
pressure with the pure gas (container, C) and the
second one is kept under vacuum (receiver, R). Their
dimensions are respectively
container: internal diameter 0.062 m, thickness
0.03 m
receiver: internal diameter 0.197m, thickness
0.003 m.
C is connected on one side to either a storage vessel
or a vacuum pump, and is kept isolated from the tube
by a computer-driven electro-valve on the opposite
side. The tube is directly opened to R, the latter being
separated by a manual valve from a precision gauge,
on which the equilibrium pressure is read after transfer. The tube internal and external diameters are respectively 1 and 1.5 mm, with a roughness of 5 pro,
typical of a drawn tube (Hodge et al., 1989).
Absolute pressure transducers are used to give direct
measurements in each reservoir. They respectively

Receiver

Precision
gauge

~,e

aE

Vacuum

Fig. 4. Schematic illustration of the experimental apparatus.

300

S. CHARTON et al.

range from 0 to 4 and 0 to 60 MPa, with an accuracy


o f l % of the full-scale length. Their output signals
are linearized and amplified. The whole chain time
constant is estimated as 0.1 s. Piezoelectric sensors
connected to fast charge amplifiers are also used to
quantify the rapidity of the pressure variations.
Fast transient temperature measurements are more
difficult to implement. The dynamic characteristic
of a thermocouple is indeed strongly dependent on
the heat transfer kinetics between the fluid and
the junction. In recently published papers, time constants as low as 3.4 ms have been reached with microthermocouples (Beckman et al., 1993). In this study,
we have used commercial chromel-alumel thermocouples. The 1 m m diameter isolation sheath is
waisted to 0.5 m m over 2 cm from the external wires
junction (in contact with the fluid). These features lead
to a time constant of 0.4 s for gas-phase measurements. The output from the container sensor is sent
through a fast amplifier and is converted afterwards.
The receiver thermocopule signals are directly ampli-

fled and converted, providing a total time constant of


the order of 0.8 s. Data are collected with an AD-2000
acquisition board.
Several experimental conditions were investigated
in order to test the model reliability: initial pressures
ranging from 10 to 45 MPa, total tube length (from
the container orifice to the receiver) of 0.3 and 1.1 m.
Furthermore, since the ratio of the heat capacities
appears to be an important parameter of the transonic flow involved, helium (7 = 1.67) and deuterium
(y = 1.41) gases were chosen for their close mole
weights. The experimental conditions are summarized
in Table 1, where experiments I - V I I I refer to helium,
and experiments I X - X V I to deuterium.
3.2. Data processin#
For all the experiments, mass values in each reservoir result from calculations based on experimental
data of temperature and pressure. The overall mass
balance of the system which is deduced points out the
influence of the sensors dynamics on the physical

Table 1. Experimental conditions investigated


Gas
nature
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII
XIV
XV
XVI

He
He
He
He
He
He
He
He
D2
D2

D2
D2
D2
D2
D2
D2

Tube
length
L (m)

Container initial
pressure (MPa)

Mean absolute
pressure deviation
(MPa)

Discharge
duration
~90 (s)

1.1
1.1
1.1
1.1
0.3
0.3
0.3
0.3
1.1
1.1
1.1
1.1
0.3
0.3
0.3
0.3

10.0
22.0
31.9
45.0
9.9
22.0
31.8
43.1
9.6
20.4
31.0
42.4
10.7
20.1
31.0
40.8

0.452
0.162
0.954
0.208
0.397
0.482
0.844
0.849
0.236
0.410
0.421
0.512
0.967
0.963
1.350
0.926

2.30
2.15
2.00
1.95
1.25
1.15
1.10
1.05
2.55
2.45
2.40
2.30
1.40
1.35
1.32
1.30

18
16

i~

14

12
i0

2
0

0.5

1.5

2.5

3.5

t i m e (s)
Fig. 5. Apparent mass fluctuation during experiment III.

4.5

A simplifiedmodel for real gas expansion


meaning of the measurements. For instance, the apparent gas mass variation for the whole system, calculated for experiment III, is plotted in Fig. 5. From
the experimental values of temperature and pressure,
it would be concluded that the amount of gas enclosed
in the hermetically sealed apparatus does not remain
constant during the transfer. This apparent mass vari-

301

ation, noticeable in the first seconds of each experiment (and attaining sometimes 20% in the case of the
short tube experiments), is of course inconsistent. It is
attributed to the dynamics of the sensors.
Obviously, the time scales of the responses of the
sensors, especially thermocouples, are of the same
order of magnitude as the time scales of the pressure

Pressure (MPa)
35
30

25

20

model
15

''~

",,~

10

'

corrected model
ex )eriment

It--

0.5

1.5

2.5

3.5
4
time (s)

Fig. 6. Pressure evolution in the container during helium transfer (experiment III).

Pressure (MPa)
0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.5

1.5

2.5

3.51m e (s)4

Fig. 7. Pressure evolution in the receiver during helium transfer (experiment III).
CE$ 51-2-J

302

et

S. CHARTON

and temperature variations during the transfer. Consequently, the temporal characteristics of the sensors
must be taken into account for a correct model validation.
The time responses of the temperature and pressure
measurement chains are represented by first-order
transfer functions of gain unity, with respective time

al.

constants given earlier. U n d e r these conditions, model


predictions can be corrected by convoluting the
simulated profiles by the inverse of the sensor transfer
function.
Since a first amplifier is coupled to the container
thermocouple, we found it more meaningful to proceed in the reverse order. The heat transfer between

Temperature (K)
313.15
293.15'

b"

2~.15

'.

253.15

model

':

233.15

c4~L~__~ measuremenls

r a w me~m~tn~t~

213.15

b.

193.15

O..

.*"

"".
173.15

"',D
"- " ' .

.~.',t'"

[]
"''-,.

"DO

~h
...o-''"

[]
.........

iC~.._- "

[]
.'"~

153.15
133.15
0

0.5

1.5

2.5

3.5

time (s)
Fig. 8. Temperature evolution in the container during helium transfer (experiment III).

Temperature (K)

,''1

"
423.15

model
.

......

' ".
398.15

348.15
323.15

corrected model
experiment

~b
ah~w~lqmu

,-..
......

i. . . . . . . .

273.15
0

0.5

1.5

2.5

3.5

time (s)
Fig. 9. Temperature evolution in the receiver during helium transfer (experiment III).

A simplified model for real gas expansion


the fluid ( f ) and the thermocouple junction (j) is
indeed expressed by

303

Equation (20) can be rewritten as a simple firstorder differential equation, introducing the time constant j of the thermocouple:

(20)

Tj = - Ajhj_f(Tj - Tf)
( p V C v ) j dd-T

dT~ =
dt

where h~_[ is the convection heat transfer coefficient


at the junction surface.

( T ~ - Tj-)
zj

(21)

Pressure (MPa)
35

20

.......

model
corrected model
expenment

10

....
0

0.5

1.5

2.5

- ---3

"
3.5

4
t i m e (s)

Fig. 10. Pressure evolution in the container during deuterium transfer (experiment XI).
Pressure (MPa)

0.6

0.5

o.,
.,

;- ~/

0.3

model
0.2

;"

corrected model
experiment

0.1

0
0

0.5

1.5

2.5

3.5

4
t i m e (s)

Fig. l l. Pressure evolution in the receiver during deuterium transfer (experiment XI).

304

S. CHARTON

The obtention of corrected experimental temperature T I is achieved by solving eq. (21) from the
junction temperatures measurements in the container.

et al.

the temperature profiles in the pressure vessel, Figs


8 and 12, is characteristic of the competition between
the two thermal phenomena occurring during the
discharge:

4. RESULTS AND DISCUSSION

The results from each experiment have been compared with the corresponding predicted data from the
model. The mean absolute deviation between the experimental and the corrected simulated pressure profiles has been estimated in each case (Table 1). Those
small departures are mainly attributed to the transducer uncertainty which is about 0.6 MPa. Two representative comparisons, III and XI (referred to as
A and B), are discussed in the following.
The pressure variations with respect to time in both
reservoirs are respectively illustrated in Figs 6, 10 and
7, 11. The raw evolutions predicted by the model have
been corrected by introduction of the sensor dynamics
as indicated earlier, and the resulting curves, labelled
as corrected model, can thus be compared with the
measured profiles. The same convention is used for
the corresponding temperature histories, Figs 8, 12
and 9, 13, except for the container, where the raw
thermocouple output signals have been corrected according to eq. (21). Therefore, in Figs 8 and 12, the
simulated temperature profiles are compared with the
real temperature evolutions in the container.
After the valve opening, the gas rapidly expands
from the container to the receiver. The pressure
balancing goes on afterwards, together with the
achievement of the thermal equilibrium. The shape of

the rapid cooling, due to the important depressurization, strongly dominates in the first
second,
after a while, the endothermic process becoming
less significant, the gas is slowly warmed up by
the surroundings.
In the receiver, conversely, a temperature increase
resulting from both compression and kinetic energy
conversion into internal energy is first observed. Then,
these exothermic processes are counterbalanced by
the heat losses through the metallic shell (Figs 9 and
13).
This qualitative behavior of the system appears to
be correctly described by our simplified model. Furthermore, there is clearly a very good agreement between the predicted profiles and the measured ones in
the container:
The departures on pressure are of the same order
of magnitude as the transducer uncertainty
(Figs 6 and 10);
the mean bulk temperatures predicted in this
vessel are very close to the corrected experimental values (Figs 8 and 12), the latter being
measured locally in the vessel.

T e m p e r a t u r e (K)

293.15 ~
273.15 - - Q

- -

253.15

r-~

233.15

[]

[]

"

model

corrected measurements

raw measurements

"1%
%.

213.15

~...

[]

193.15

. o~I

6 .... b'-'-.

0...<2...~-~-"

:~

i1

173.15
153.15
133.15
0

0.5

1.5

2.5

3.5

4
t i m e (s)

Fig. 12. Temperature evolution in the container during deuterium transfer (experiment XI).

A simplified model for real gas expansion

305

Temperature (K)
423.15

398.15

373.15 ~

348,15

0.5

1.5

2.5

3.5

time (s)
Fig. 13. Temperature evolution in the receiver during deuterium transfer (experiment XI).
In the receiver, however, the model obviously
overestimates the thermal variations: a sharp temperature increase is predicted in the first milliseconds,
which is hardly compensated by the convective heattransfer. The measurements on the other hand reveal
a smoother variation and lower temperature values
(Figs 9 and 13). Too high predicted pressures follow
from this temperature departure as shown in Figs
7 and 11.
The weaker agreement observed in the receiver may
be due to some inconsistencies in the assumptions
made during the model development. Among them,
the less reliable are the following:
spatially uniform temperatures
stagnant conditions
natural convection dominance.
Indeed, because the receiver has finite dimensions,
the sonic gas jet is likely to be reflected by the walls
and then deviated by the continuous incoming gas
flow. Therefore, nonstagnant conditions are prevailing in the enclosure and prevent the fluid kinetic
energy from being entirely and instantaneously restituted in thermal energy. Sharp velocity fluctuations
are taking place, thus generating a complex flow pattern in this enclosure, and strongly promoting forced
convection because of the resulting buoyancy. Conversely, the velocity of the gas at the pipe entrance is
considerably less significant than that of the exhausting sonic jet; therefore the assumptions of gas stagnation and natural convection dominance in the container
are justified.

To take into account the influence of forced convection in the receiver, a different correlation for Nusselt
number
Nu = 0.023 Re 8 Pr aa

(22)

has been introduced in the model instead of eq. (1).


The Reynolds number has been given empirically, as
there is no simple way to obtain a mean value of it.
Two values of Re were used: 104 and 105 . When Re
increases, the temperature decrease is stronger and the
curves obtained with this model (Figs 14 and 15) are
situated closer to the experimental curves and well
under the natural convection curve. Unfortunately,
the main problem with the model including forced
convection is that the Re number cannot be predicted
in the frame of a simplified approach; consequently
this model cannot be considered as totally predictive.
The particularly important discrepancy between
raw experimental and corrected simulated temperatures may also be due to the significant time constant
(around 0.8 s) of the corresponding measurement
chain. The precision of the correction brought by the
convolution process in this case may be insufficient.
Lastly, the temperature field in the downstream
reservoir is obviously very complex and strongly dependent on time (through the relative positions of the
incident and reflected gas jets). It must be kept in mind
that in this work we compare a mean value of the
temperature (uniform temperature predicted by the
simplified model) to a local in situ measurement. For
the reasons evoked above, this procedure is appropriated in the discharging vessel, but the comparison

306

S. CHARTON et al.

Pressure (MPa)
0.7

0.6

0.5

'

0.4

..o~o

.dlo

, , 4~'

j;

0.3

0.2

--

."1

0,j

-.

natural convection
forced convection - Re=le4
forced convection - Re=le5
ex ~eriment

0
0

0.5

1.5

2.5

3.5
4
t i m e (s)

Fig. 14. Pressure evolution in the receiver during helium transfer (experiment III)--influence of forced
convection heat transfer.

Temperature (K)
473.15

448.15
423.15

natural convection
forced convection - Re=le4
forced convection - Re=le5
experiment

......

39& 15

373.15

323|5

f ~

!1,

/6

It

/,;,~

41.o

"

.... a ....... !; ...... o ...... o

o./,

lw

, - - I ~ a m

oo

273.15

0.5

1.5

2.5

3.5

4
t i m e (s)

Fig. 15. Temperature evolution in the receiver during helium transfer (experiment III)--influence of forced
convection heat transfer.

is n o t so representative in the receiver. However, in


the receiver, the pressure evolution is correctly described by the model, a n d the transfer d u r a t i o n as well
as the final pressure are accurately predicted
(Figs 2 a n d 7).

Since the model simulates perfectly the p h e n o m e n a


occurring in the container, the predicted discharge
d u r a t i o n Z9o, defined as the time required to decrease
the initially stored mass of gas of 90%, h a s been
plotted with respect to the initial pressure, in o r d e r to

A simplified m o d e l for real gas e x p a n s i o n

307

Discharge Duration "c~(s)


3

He- L=l.lm
D2 - L = I . I m

2.5
-0 ...................................
....... 0

---..~.

""A

1.5
P. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

-0

,.o

.......A-.........................
............ .A

0.5

t,

He - L = 0 . 3 m

D2 - L = 0 . 3 m

10

30
40
50
Container Initial Pressure (MPa)

20

Fig. 16. V a r i a t i o n s of the d i s c h a r g e d u r a t i o n Zgo with the initial s t o r a g e pressure.

illustrate the properties of the frictional flow (Fig. 16).


For a given gas and a given tube length, %0 decreases
as the storage pressure increases. The latter is indeed
the driving force of the discharge. It appears in Fig. 16
that %0 is also strongly dependent on the tube length
and the gas molarity:
(a) According to the Fanno flow characteristics, the
transfer time is nearly doubled when the tube length is
increased from 0.3 to 1.1 m (all other things being
equal). The longer tube inducing a higher frictional
pressure drop and hence a higher acceleration of the
subsonic flow, the thermodynamic consistency requires the inlet gas flow rate to be reduced in a higher
proportion for this long tube than for the shorter one.
(b) For a given initial pressure and a fixed tube
length, the gas transfer between the two reservoirs is
shown to be slower in the case of deuterium than
when helium is concerned. Keeping in mind that most
of the gas is blown out under sonic conditions, the gas
velocity at the pipe outlet is equal to the sonic velocity
c = x / - ~ ( R / M ) T . Thus, for a given temperature, the
deuterium over helium velocity ratio is given by the
square root of their ?-ratio. Figure 16 reflects that the
mean value of the helium over deuterium discharge
duration is effectively close to the reciprocal of this
ratio:

= 0.92

while

= 0.89.

a thin tube. The flow is assumed to be one-dimensional and quasi-steady. Real gas effects and heat
transfer phenomena are neglected in the tube but
taken into account in the reservoirs. Finally, providing sensible assumptions, the fast transient flow of the
compressible fluid can be solved with minimal calculation efforts, simply by setting the system geometry
and the fluid properties.
Besides the apparent difficulty of the simulation
task, the experimental investigation of the process
appeared to be critical. Owing to the fact that the
major amount of gas is blown in less than half a second, the measurements accuracy is strongly dependent on the dynamic characteristics of the sensors
involved. The actual pressure and temperature evolutions can be estimated anyway by taking into account
the transfer function of the measurement chains. The
convolution product of the predicted data by the
inverse of this function leads to a great improvement
in the meaning of the comparison between simulations results and experimental data.
The model predictions have been compared with
several experimental data. The qualitative behavior of
the process appears to be very well described and
a very close quantitative agreement is reached in the
discharging vessel. A weaker agreement is reached in
the downstream reservoir probably because of the
stagnant bulk approach derived in the model. However, the receiver conditions have a feeble influence on
both the container discharge and the choked flow in
the pipe which are correctly described.

5. C O N C L U S I O N

A simplified model has been proposed for simulating gas transfer between two enclosures through

NOTATION

c
C/

velocity of sound, m/s


friction factor

S. CHARTONet al.

308
Cp, C~

D
#
h
hw
L
M
P
R

S
t

T
u

U
V
x
Z

specific heat, J/(kg K)


tube diameter, m
gravity, m/s 2
specific enthalpy, J/kg
heat transfer coefficient, W/(m 2 K)
tube length, m
mole weight, kg/mol
Mach n u m b e r
absolute pressure, Pa
universal ideal gas constant, J/(K mol)
area of the heat exchange surface, m 2
time, s
absolute temperature, K
velocity, m/s
internal energy, J
volume, m 3
abscissa
compressibility factor

Greek letters
isobaric dilatation coefficient, K - 1
ratio of specific heats
7
tube roughness, m
8
2
thermal conductivity, W / m
/z
viscosity, kg/(m s)
density,
kg/m a
P
characteristic time
tube section, m 2
fZ
Subscripts
f
fluid
j
junction
w
wall
oo
bulk
Exponents
0
stagnation property

choking condition
steady condition
REFERENCES

Beckman, P., Roy, R. P., Whitfield, K. and Hasan, A., 1993,


A fast-response microthermocouple. Rev. Sci. lnstrum. 64,
2947-2951.
Cess, R. D., 1973, Free-convection boundary-layer heat
transfer, in Handbook of Heat Transfer (Edited by W. M.
Rohsenow and J. P. Harnett), Section 6, p. 613. McGrawHill, New York.
Fletcher, C. A. J., 1991, Computational Techniques for Fluid
Dynamics, Vol. 2: Specific Techniques for Different Flow
Cate#ories, 2nd Edition. Springer, Berlin.
Haque, M. A., Richardson, S. M. and Saville, G., 1992,
Blowdown of pressure vessels. I. Computer model. Trans.
Instn Chem. Engrs 70, 3-9.
Hodge, B. K., Taylor, R. P. and Coleman, H. W., 1989,
Predicting turbulent rough-wall skin friction and heat
transfer, in Encyclopedia of Fluid Mechanics, Vol. 8: Aerodynamics and Compressible Flows (Edited by N. P.
Cheremisinof), Chap. 13. Gulf Publishing Company,
Houston.
Issa, R. I. and Spalding, D. B., 1972, Unsteady one-dimensional compressible frictional flow with heat transfer. J.
Mech. Engng Sci. 14, 365-369.
Leung, J. C., 1986, Simplified vent sizing equations for
emergency relief requirements in reactors and storage
vessels. A.I.Ch.E.J. 32, 1622-1634.
Levenspiel, O., 1977, The discharge of gases from a reservoir
through a pipe. A.I.Ch.E.J. 23, 402-403.
Reid, C., Prausnitz, J. M. and Poling, B. E., 1987, The
Properties of Gases and Liquids, 4th Edition. McGrawHill, New York.
Shapiro, A. H., 1953, The Dynamics and Thermodynamics of
Compressible Fluid Flow, Vol. I. The Ronald Press Company, New York.
Soave, G., 1980, Rigorous and simplified procedures for
determining the pure-component parameters in the Redlich-Kwong-Soave equation of state. Chem. En#ng Sci. 35,
1725-1734.
Xia, J. L., Smith, B. L. and Yadigaroglu, G., 1993, A simplified model for depressurization of gas-filled pressure
vessels. Int. Comm. Heat Mass Transfer 20, 653-664.

Você também pode gostar