Ac 2007-2695: Modeling Compressible Air Flow in A Charging or Discharging Vessel and Assessment of Polytropic Exponent
Ac 2007-2695: Modeling Compressible Air Flow in A Charging or Discharging Vessel and Assessment of Polytropic Exponent
Ac 2007-2695: Modeling Compressible Air Flow in A Charging or Discharging Vessel and Assessment of Polytropic Exponent
Page 12.1075.1
Abstract
In this work, the classic problem of charging and discharging of a pressurized tank is
studied. This experiment allows students to gain a deeper understanding of polytropic processes
and compressible flows. The experiment apparatus described in this study allows for direct
measurement of the pressure and temperature within the tank, and utilizes a LabView based
computerized data acquisition system. To assure accurate measurements of these parameters, a
fast-response thermocouple and a high accuracy variable reluctance pressure transducer is
employed.
A model was developed to predict the pressure and temperature of the air in the tank
during charging and discharging. The model incorporates compressible flow in both sonic and
subsonic flow regimes, and models the air as undergoing a general polytropic process. The
model was compared with experimental data to empirically determine the polytropic exponent.
The values of polytropic exponent obtained through the phenomenological model were
compared to those determined by a graphical technique to determine to polytropic exponent.
Results show that the polytropic exponent varies with initial pressure and throat area, as well as
with time. Thus a constant value for polytropic exponent generally yields an unsatisfactory
prediction for temperature and pressure. It is found that a discharge coefficient must be included
in the analysis to accurately match the data, due to frictional effects through the throat. Further,
the experiment also indicates that heat transfer through the vessel walls plays a major role in the
process.
Introduction
The analysis of a pressurized air tank being charged or discharged is one of the most
common applications of compressible flow presented in undergraduate fluid mechanics courses.
The scenario usually involves an initially pressurized vessel which is suddenly open to a lower
outside pressure (such as atmosphere) through a small opening. The goal of this experiment is to
predict either the time required to discharge the tank, or the pressure inside the tank, after a
specified time. The exercise is useful to students because it is a rather straightforward
application of conservation of mass, and introduces the concepts of choked and subsonic flows.
Further, the solution integrates aspects of thermodynamics and heat transfer, making for an
excellent capstone experiment in thermal sciences.
The most comprehensive solution to the problem is presented by Bober et al.1 They
applied conservation of energy to a discharging tank of air to predict the temperature and
pressure inside the tank as a function of time. They analyzed both choked-flow and subsonic
regimes, and incorporated the heat transfer through the walls of the tank. The authors modeled
the flow through the exit nozzle as isentropic, and the heat transfer as natural convection at the
inside and outside wall surfaces. They approximated the heat transfer through the wall as quasi-
Page 12.1075.2
steady-state. In spite of these simplifications, excellent agreement was found between the model
and experimental data. However, the solution itself is complicated, involving the application of
conservation of mass to the air, and conservation of energy to the air and the tank wall. The
result is three ordinary differential equations to be solved simultaneously for the temperature of
the air, the mass of the air, and the temperature of the inside wall of the tank.
A second, simpler approach to the problem avoids modeling the heat transfer explicitly,
and instead treats the air inside the tank as undergoing a polytropic process (pvn = constant). The
solution to the problem involves applying conservation of mass to the air inside the tank. This
solution has recently been demonstrated by Dutton and Coverdill2. They modeled the pressure
response of air tanks during either charging from another pressurized tank, or discharging to
atmosphere. Transient temperature response was not recorded, but both choked and subsonic
regimes were modeled. Two tank volumes were studied, with different nozzle sizes. The
authors further limited their analysis to two polytropic processes: isentropic (n = 1.4) and
isothermal (n = 1). They demonstrated that discharging a larger tank (relatively slow discharge
process) approached isothermal behavior, while discharging a small tank (relatively fast
discharge) approached adiabatic behavior. In contrast, the charging of either tank volume was
best approximated by the adiabatic model.
The work of Dutton and Coverdill2 was limited to isothermal and isentropic processes,
which allow for closed-form solution. Although these processes bound the solution, their model
does not solve for a general polytropic process. Unfortunately, when modeling the air expansion
or compression as a polytropic process, the polytropic exponent n is not typically known a priori.
In fact, determining the polytropic exponent is itself a worthy experiment for an undergraduate
laboratory.
In this work, the authors develop a model for predicting the pressure and temperature of
air charging into or discharging out of a tank. The model incorporates compressible flow in both
choked and subsonic flow, and models the air as a general polytropic process of power n. The
model is applied to two scenarios: (a) an initially pressurized tank discharged to atmosphere, and
(b) an initially evacuated tank being charged from atmosphere. The polytropic exponent, n, is
found empirically by matching the model to the data. The results of this method are compared
to the graphical technique for calculating the polytropic exponent. The difficulties in
determining this exponent, and the challenges of modeling the system without advance
knowledge of n, makes this experiment challenging for the undergraduate laboratory, and a
Page 12.1075.3
pedagogical approach is suggested, along with alternate approaches to facilitate student learning
outcomes.
Experimental Facility
A schematic diagram of the tank is presented in Figure 1. The tank is constructed from
20 cm ID (approximately 8 inch) schedule 40 steel pipe, with welded end plates, and constructed
with a gasket seal in the middle to allow access to the inside. The tank is pressurized from a
compressed air source through an inlet valve. Alternately, a vacuum pump is connected to the
valve to evacuate the tank. The tank discharges to or charges from the atmosphere via one of
three ports, connected to the opposite end of the tank and controlled by ball valves. The exterior
end of each valve is connected to a threaded plug, with a hole machined through it to form a
throat. Each throat diameter was chosen to ensure that the throat area is the smallest area in the
exit line; that is, to ensure that the throat is where the flow will be choked.
The pressure inside the tank is measured using a Validyne DP-15 variable reluctance
pressure transducer, which is connected via a carrier-demodulator to a LabView-based PC data
acquisition system. The transducer is calibrated in the laboratory to within ± 3 kPa
(approximately 0.05% of full scale). Temperature is measured with a type T micro-
thermocouple probe made by Paul Beckman Company. The probe has a junction of
approximately 0.05 mm (0.002 in), and has an estimated time constant of 0.02s for the conditions
of this experiment. The thermocouple, accurate to ± 1ºC, is shielded using a perforated mylar
tube, which was constructed to allow air to pass freely but shield the junction fro thermal
radiation. A photograph of the experimental facility is presented in Figure 2.
Variable reluctance
pressure transducer
Pressure relief
to DAQ
valve
P
Discharge
valves
51.5 cm
P
Air in 20 cm
pv n ? pt / n = constant , (1)
where p is pressure, v is specific volume, and is density, can be combined with the Ideal Gas
Law, pv ? RT , to obtain
pT n /( n /1) = constant . (2)
This relationship can be linearized by taking the natural logarithm of both sides of the equation,
producing
n
ln p - ln T =constant . (3)
n /1
Thus, plotting the natural logarithm of the experimentally obtained pressure and temperature
(ln(p) against ln(T)) should yield a linear relationship, with slope equal to n/(n-1).
An alternate derivation is to evaluate Eqn. (2) between two states 1 and 2 to obtain
Page 12.1075.5
n /( n /1)
p 2 Ã T2 Ô
?Ä Õ . (4)
p1 ÄÅ T1 ÕÖ
n ln( p 2 / p1 )
? , (5)
n / 1 ln(T2 / T1 )
n ln( p 2 ) / ln( p1 )
? . (6)
n / 1 ln(T2 ) / ln(T1 )
Equation (6) again demonstrates that the term n/(n-1) represents the slope of the relationship
between ln(p) and ln(T).
A model for the charging or discharging process of a vessel filled with air is applied to
the control volumes depicted in Fig. 3. The air is assumed to behave as an ideal gas, and the
flow through the thoat (exit, in the case of discharging tank) will be approximated as isentropic.
However, because of the possibility of heat transfer to or from the surroundings, the air inside the
tank is assumed to undergo a more general polytropic process, as shown in Eqn. (1). The model
will also allow for the polytropic constant n to vary with time, as will be considered later in this
work. Allowing for a general, time-varying polytropic exponent n precludes a closed-form
solution; therefore a numerical solution is developed.
Throat Throat
t CV ? t 0 At , Vt t atm ? t 0
At , Vt
TCV ? T0 patm t CV , TCV , pCV Tatm ? T0
pCV ? p0 p atm ? p0
dmCV
? ‒ m% t , (7)
dt
where mCV is the mass inside a tank of volume V , which can be expressed as tV . The term
m% t refers to the mass flow rate at the throat; a positive value represents a charging process, and a
negative value represents a discharging process. Assuming uniform properties within the tank, a
rigid control volume, and uniform properties along the throat surface, Eqn. (7) can be written as
dt CV t AV
?‒ t t t . (8)
dt V
For either discharging or charging process, the assumption if isentropic flow through the throat
gives the familiar relation between density at the throat to stagnation conditions6,
/1 /( i /1)
Ç i /1
t t ? t 0 È1 - M e2 Ù , (9)
É 2 Ú
where i is the specific heat ratio for isentropic flow of air, and M e is the Mach number at the
throat. In the above, t 0 is the stagnation density; in the discharging process, the air inside the
tank is assumed to be at stagnation conditions. The velocity at the throat is given by
Vt ? M t *iRTt +
1/ 2
, (10)
with R being the gas constant for air. The absolute temperature at the throat Tt is related to
stagnation conditions by
/1
à i /1 2 Ô
Tt ? T0 Ä1 / Mt Õ . (11)
Å 2 Ö
i. Choked flow ( p atm / pCV > 0.528 for discharging process, pCV / p atm > 0.528 for
charging):
M t ,i ? 1
ii. Subsonic Flow ( p atm / pCV @ 0.528 for discharging process, pCV / p atm @ 0.528 for
charging):
1/ 2
ÊÍ 2 ÇÃ p Ô (i /1) / i ÛÍ
M t ,i ? Ë ÈÄÄ 0,i ÕÕ / 1Ù Ü , (14)
ÍÌ i / 1 ÈÉÅ pb Ö ÙÚ ÍÝ
where pb ? patm for the discharging process, and pb ? pCV ,i for the charging process.
Once the tank air density at time step i+1 has been computed using Eqn. (13), the polytropic
relationship, Eqn. (1), is modified to determine the tank pressure at that time step,
ni
Ãt Ô
pCV ,i -1 ? pCV ,i ÄÄ CV ,i -1 ÕÕ , (15)
Å t CV ,i Ö
where ni is the instantaneous value of the polytropic exponent, allowed to vary with time in the
model. Finally, the tank temperature T0,i -1 can be predicted using the Ideal Gas Law,
pCV ,i -1
TCV ,i -1 ? . (16)
t CV ,i -1 R
Results
Figure 4 depicts pressure and temperature histories obtained from the facility for a variety
of throat sizes and initial pressures. Figure 4(a) compares pressures and temperatures measured
for an initial pressure of 790 kPa abs (100 psig), with different exit diameters: dt = 1, 2.1, and
2.71 mm. As expected, the pressure decreases more rapidly for a larger exit diameter.
Correspondingly, the temperature drops more rapidly, and to a lower minimum value. The
presence of heat transfer is also indicated by the temperature data of the 2.1 mm and 2.71 mm
exits, which show a rapid temperature recovery after the tank has discharged. Moreover, the
slope of the temperature during recovery is higher for the 2.71 mm exit. That same condition
also yields a lower minimum temperature, which would result in a higher potential for heat
transfer.
Page 12.1075.8
Figure 4(b) compares pressure and temperature plots for the 1 mm throat for three initial
pressures, 790 kPa abs (100 psig), 514 kPa abs (60 psig), and 307 kPa abs (30 psig). Again, the
tank takes longer to empty when pressurized to a higher initial pressure. In addition, the rate of
pressure decrease is higher, which is consistent with the fact that higher tank pressures yield
higher exit mass flow rates. Interestingly, the temperature histories are very similar for all three
conditions, having nearly identical initial slopes, and reaching similar minimum temperatures.
Finally, Figure 4(c) compares the pressures and temperatures measured during a charging
process. As before, the pressure data show the expected trend that a tank with a smaller throat
takes longer to pressurize. The temperature data also show that a larger throat results in a larger,
and faster, rise in tank temperature. Again, the data for the 1 mm throat show such a slow
charging process that the temperature reaches a maximum, and then begins recovering before the
end of the charging process.
900 900 120
800 800
p init = 790 kPa 100
Pressure (kPa absolute)
Temperature (K)
Temperature (K)
310
280 280 d t = 1 mm
275 275 514 kPa 305
270 270 790 kPa
2.1 mm
265 300
265
260 260
295
255 2.71 mm 255
250 250 290
0 50 100 150 200 0 50 100 150 200 0 50 100 150
time (s) time (s) time (s)
(a) pinit = 790 kPa abs (b) dt = 1 mm (c) pinit = 3.7 kPa abs
Figure 4. Pressure and temperature profiles for selected experimental conditions: (a)
varying throat diameter at initial pressure 790 kPa absolute (100 psig); (b) varying
initial pressure for throat diameter 0.1 mm; (c) varying throat diameter at initial
pressure 3.7 kPa absolute (-14 psig).
Page 12.1075.9
Graphical Method for Obtaining Polytropic Exponent
Figure 5 illustrates the graphical method of obtaining the polytropic exponent for an
initial pressure of 790 kPa abs (100 psig), and a throat diameter of dt = 2.1 mm. The plot of ln(p)
as a function of ln(T) is fairly linear during the first 15 seconds of the event, or approximately the
first quarter of the discharge time. For all the discharging processes studied, a linear curve fit
was typically appropriate for the first quarter to the first half of the event. The deviation from
linear behavior in these plots suggests that the polytropic exponent n varies with time. This
conclusion seems reasonable, since the air temperature varies greatly during discharge, which
would affect the heat transfer. In contrast, the plots of ln(p) as a function of ln(T) for the
charging events remain fairly linear throughout the entire process, suggesting that the polytropic
exponent, and the influence of heat transfer, is fairly constant.
On the other hand, the measured polytropic exponent was not affected significantly by
throat size when the tank was initially evacuated. For all throats, the polytropic exponent was
close to an isothermal value, approximately 1.01 to 1.02. In fact, the temperature rise during
charging is small – on the order of about 20 K. The reason for the nearly isothermal behavior is
likely due to the fact that the air entering the tank is coming from constant temperature, ambient
conditions.
A trend that particularly interesting is that, for a tank discharging through a particular
throat, n appears to increase with initial pressure. This trend seems unexpected, since higher
initial pressures yield higher mass flow rates, which should result in lower system temperatures.
In turn, the lower temperatures would promote heat transfer, which should make the polytropic
exponent decrease. However, Johnson et al.7 show that the discharge coefficient for sonic
nozzles increases with increasing back pressure. Higher discharge coefficients correspond to
higher flow efficiency, which tends toward isentropic flow. This may explain the trend seen in
the measured polytropic exponent with increasing initial pressure.
Phenomenological Model
The results of the previous discussion reveal that the polytropic exponent is not known a
priori, and from a student’s perspective is difficult to predict. One way to approach this
difficulty is to empirically choose the polytropic constant which best matches the model to the
experimental data.
Page 12.1075.10
The phenomenological model is compared with experimental data for a throat diameter of
2.1 mm, discharging from an initial pressure of 790 kPa (100 psig), in Figure 7. The model was
performed over a range of polytropic exponents, including n = 1.17, the value obtained using the
graphical method. Initially, both the pressure and temperature data are modeled closely for n =
1.17. This is consistent with the graphical technique for determining n, which was also effective
only during the initial portion of the event. Thereafter, the model predicts a more rapid decline
in pressure than measured. In contrast, the model predicts temperatures that are significantly
lower than measured.
14.0
13.5 slope
n
=
13.0
n /1
ln(p)
12.5
12.0
11.5
11.0
5.55 5.60 5.65 5.70 5.75
ln(T)
Figure 5. Graphical method for obtaining polytropic exponent n during initial tank
discharge, for initial pressure 790 kPa abs (100 psig), dt = 2.1 mm. The linear portion
of the plot spans approximately the first 15 seconds of discharge; the slope shown
yields n B 1.17.
Page 12.1075.11
1.20
1.18
1.16
1.14
Polytropic exponent n
1.12
1.02
1.00
0 p atm 200 400 600 800 1000
Initial Pressure (kPa absolute)
Examining the model results obtained for various values of n, it is observed that the
experimental pressure curve is more closely fit by an n = 1.17 model initially, but the isothermal
model appears to match the experimental data more closely near the end of the process.
Although the modeled pressure is only slightly sensitive to the value of n, the temperature
prediction is extremely sensitive to n, and it appears that no single value of n satisfactorily
describes the experimental temperature.
Page 12.1075.12
900 300
n = 1.00
800 290
Exp. Data
700 Model 280
Pressure (kPa absolute)
270
600
Temperature (K)
260
500 n = 1.00
250 n = 1.10
400
240
n = 1.10
300
230 n = 1.15
200 n = 1.17
220
n = 1.20 Exp. Data (graphical solution)
100 210 n = 1.20
Model
0 200
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Time (s) Time (s)
Figure 7. Comparison of model with experimental data for pinit = 790 kPa absolute
(100 psig) and de = 2.1 mm. Includes parametric study of the effect of varying
polytropic constant.
One explanation for the discrepancy between the predicted and measured temperatures is
that the relative influence of heat transfer changes with time, since the temperature of the air in
the vessel varies. This effect would make the polytropic exponent a transitory value. To gain
insight into this effect, the data obtained for pinit = 790 kPa (100 psig) and dt = 2.1 mm is
reexamined. In Figure 8, the graphical method is used to calculate n, but this time n is found
from the instantaneous slope of ln(p) vs. ln(T) data. The result is a measure of n as a function of
time, and Figure 8 shows the value of n decreases. Indeed, if this time-varying polytropic
exponent is incorporated into the model, the resulting semi-empirical model more closely
approximates the behavior of the system. This result is demonstrated in Figure 9.
The semi-empirical model, with varying polytropic exponent, still under-predicts the
pressure and temperature response slightly. This discrepancy could be due to the effects of
friction. Following Johnson7, a discharge coefficient could be incorporated into the mass flow
rate; discharge coefficients for airflow through sonic nozzles range from approximately 0.94 to
0.97, and increases with Reynolds number. To examine whether friction alone could explain the
discrepancy in the model, a discharge coefficient was added to the modified model. A discharge
coefficient value was selected that best matched the experimental data. A value of 0.92
produced excellent results, suggesting that the presence of friction could explain the discrepancy
between the modified model and the data.
Page 12.1075.13
1.4
experimentally determined
1.3 Curve Fit
Polytropic constant, n
1.2
1.1
1.0
0.9
0.8
0 10 20 30 40 50 60
Time (s)
Finally, the model is applied to the charging process, illustrated in Figure 10. In the
charging scenarios, a constant value on n was satisfactory, with the value of n determined by the
graphical analysis of the experimental data. The use of a constant value for n is supported by a
plot of the instantaneous n value as a function of time, shown in Figure 11. The model slightly
overpredicts the pressure response, but incorporating a discharge coefficient of 0.94 provided an
excellent fit to the data. The fact that the discharge coefficient would be higher under these
conditions than the conditions of Figure 9 makes sense, since the inlet velocity during charging
from ambient pressure is lower than the exit velocity for a back pressure of 790 kPa, and hence
the effect of friction should be correspondingly lower.
Page 12.1075.14
900 300
270
600
Temperature (K)
260 varying n
500
250
varying n with discharge
400
varying n coefficient = 0.92 240
300
data 230
200 220 constant n
Exp. Data
100 210
constant n Model
0 200
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Time (s) Time (s)
120 320
Constant n = 1.014
100 Constant n = 1.014 315
Pres sure (kPa absolute)
80 310
Temperature (K)
40 300
Exp. Data
20 Model 295 Exp. Data
Model
0 290
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)
Figure 10. Comparison of model with experimental data for pinit = 3.7 kPa absolute
(-14 psig) and dt = 2.1 mm. Model incorporated a polytropic constant value of n =
1.014.
Page 12.1075.15
1.4
1.3
1.2
Polytropic Exponent, n
1.1
1.0
0.9
0.8
0.7
0 5 10 15 20 25 30
time (s)
Conclusions
In this work, a model was developed to predict the pressure and temperature of air in a
tank during charging and discharging. The model incorporates compressible flow in both choked
and subsonic flow regimes, and models the air as undergoing a general polytropic process. The
model was compared with experimental data to empirically determine the polytropic exponent.
The values of polytropic exponent obtained through the phenomenological model were
compared to those determined by a graphical technique to determine to polytropic exponent.
The experimental data, modeling effort, and comparison of techniques to determine the
polytropic exponent resulted in several observations. For the discharging processes studied in
this work:
‚ The polytropic exponent, determined using the graphical technique, was found to be a
function of operating conditions in the vessel. Larger throat diameters yielded lower
values of n, while higher initial pressures yielded higher values of n.
‚ The graphical technique for calculating the polytropic exponent based on temperature and
pressure histories was suitable only for the initial period of discharge. In fact, the
polytropic exponent was found to vary with time. This is consistent with the fact that the
polytropic exponent n is an indicator of the influence of heat transfer which, for an
uninsulated vessel, varies with time as the conditions inside the tank vary.
‚ The phenomenological model underpredicts the pressure and temperature histories when
Page 12.1075.16
a constant value of n was used in the model. The discrepancy may be explained,
however, by the variation in polytropic exponent (and hence, heat transfer) with time, as
well as the presence of flow friction.
Examining the initially evacuated vessel being charged from ambient air:
‚ The polytropic process was found to be nearly isothermal, with n being approximately
constant with time. The graphical method could therefore be used to determine n over
nearly the entire the process, not just initially.
‚ The phenomenological model matched the experimental temperature and pressure data
very well, and the small discrepancy between model and data could be explained by the
existence of flow friction.
The goal of this work is to develop an experiment and solution approach that is adaptable
to the undergraduate laboratory. The difficulty of applying this model is that the polytropic
exponent is not known, or easily determined, at the outset of the analysis. Moreover, the solution
is highly sensitive to the value of n, and since n varies with time, the model results – particularly
the temperature variation – does not satisfactorily compare with experimental data without a
detailed modeling of both heat transfer and flow friction. However, for the initially evacuated
tank being charged with ambient air, the polytropic exponent is approximately constant, and
modeling yields a pressure and temperature prediction that compares well with measurements
without a great deal of additional analysis.
Acknowledgments
The authors wish to acknowledge Donald Alman (’69) and Robert Dobbs (’78), whose
Senior Projects helped develop the hardware for this experiment as well as earlier versions. The
authors also thank Professor Emeritus Ed Stoeffel, who supervised the early development of this
Page 12.1075.17
experiment, and whose mark on the department is still felt to this day.
References
[1] Bober, W., Heinzman, J.D., and Laptchak, D.M., “Non-Steady Nozzle Flow: An Educational Experiment,” Int.
J. Mechanical Engineering Education, Vol. 7, No. 3, pp. 131-136, 1979.
[2] Dutton, J.C. and Coverdill, R.E., “Experiments to Study the Gaseous Discharge and Filling of Vessels,” Int. J.
Engineering Education, Vol. 13, No. 2, pp. 123-134, 1997.
[3] Alman, Donald, “The Emptying Process of a Pressure Vessel,” Senior Project 69-0319, California Polytechnic
State University, San Luis Obispo, 1969.
[4] Dodds, Robert, “Polytropic Blowdown Test Stand,” Senior Project 78-0545, California Polytechnic State
University, San Luis Obispo, 1978.
[5] Hawkins, G.A., Thermodynamics, John Wiley and Sons, New York, 1946.
[6] White, F.M., Fluid Mechanics, 5th Edition, McGraw Hill, New York, 2003.
[7] Johnson, A.N., Espina, P.I., Mattingly, G.E., Wright, J.D., and Merkle, C.L., “Numerical Characterization of the
Discharge Coefficient in Critical Nozzles,” NCSL Workshop and Symposium, Session 4E, pp. 407-422, 1998.
Page 12.1075.18