You are on page 1of 14

MATHEMATICAL

COMPUTER
MODELLING
PERGAMON Mathematical and Computer Modelling 38 (2003) 519-532
www.elsevier.com/locate/mcm

Model of the Filling


of an Automotive Fuel Tank
S. FACKRELL, M. MASTROIANNI* AND G. W. RANKIN+
Mechanical, Automotive and Materials Engineering, University of Windsor
Windsor, Ontario, Canada N9B 3P4
ranlcin@uwindsor.ca

(Received November 2002; revised and accepted March 2003)

Abstract-one of the-problems associated with automotive fuel tanks is premature shut-off during
the filling process. This occurs when the fuel dispensing nozzle shuts off before the fuel tank is
completely filled.
A comprehensive but simplified lumped parameter model has been developed to determine the
pressure and flow rate associated with the automotive fuel tank for the purpose of predicting pre-
mature shut-off. Unlike previous work, this model includes the entire fuel system, including a closed
rectangular tank with a filler tube, rollover valve, and vent tube that connects the filler tube to
the tank. The system was divided into several control volumes and connecting valves. Continuity
equations were written for esch of the control volumes and energy equations were written for each
of the connecting valves. The model includes the effects of the fuel tank system geometry, fuel noz-
zle dispensing rate, fuel volatility, multiphase, multicomponent flow through the filler tube, air/fuel
vapour in the tank dome, vapour generation and choking in the vent tube, and the rollover valve due
to the low speed of sound of the vapour droplet flow.
This mathematical model consists of five nonlinear, integral equations and a number of algebraic
equations. A commercial numerical solver (Simulink) was used to solve these equations subject to
appropriate boundary conditions.
The simplified model was used to generate results to compare with existing experimental results. In
spite of its simplicity, it was capable of predicting the nature of the shut-off for all of the corresponding
sets of experimental data. Prediction of the magnitudes of the tank dome pressure and the filler tube
pressure were, however, only in fair agreement with the experimental values. The model was also used
to perform a sensitivity analysis of model parameters. The results were found to be most sensitive
to vent area around the filler nozzle, liquid dispensing flowrate, and fuel volatility. @ 2003 Elsevier
Ltd. All rights reserved.

Keywords--Lumped-parameter model, Fuel tank filling, Vapour generation.

1. INTRODUCTION
One of the major sets of components in an automobile is the fuel supply system. A typical
automobile fuel system consists of four major components: the filler tube, the fuel tank, the vent
tube, and the rollover valve. A schematic of the fuel tank system is shown in Figure 1.
*Currently with Kautex Textron, Windsor, Ont.
tAuthor to whom all correspondence should be addressed.
The authors would like to thank Mr. L. Savoni and H. Atsbha of Kautex Textron for their support of this work.
Financial support was provided by Kautex Textron and the Natural Sciences and Engineering Research Council
of Canada through Collaborative Hesearch Development Grant, Number CRD 227033/99.

0895-7177/03/s - see front matter @ 2003 Elsevier Ltd. All rights reserved. Typeset by &+S-TEX
doi: lO.l016/SO895-7177(03)00248-6
520 S. FACKRELL et al.

Fuel Vapoar/Ak Mixture


Fuel
Tank

Figure 1. Schematic of fuel tank system.

Phase I Phase II 1PbaseIII ,


ti I I

Time

Figure 2. Pressure-time relationship depicting different phases during normal tank


filling process.

The filler tube receives the liquid fuel/air mixture that exits the fuel dispensing nozzle. The
fuel dispensing nozzle flow contains a small amount of air due to the air entrainment process
associated with the shut-off sensor and trip mechanism operation [l]. The liquid fuel/air mixture
is further mixed with the fuel vapour which returns to the filler tube through the return vent
tube creating a multiphase, multicomponent flow. The multiphase, multicomponent flow enters
the tank where the liquid falls to the bottom of the tank and the gases are added to the vapour
space of the tank. The return vent tube, runs along the filler tube and exits near the mouth of
the filler tube. It has an orifice that protrudes through the top of the tank. The rollover valve
is attached to the carbon canister and is designed to prevent any liquid from escaping the fuel
tank in the event of an accident (the car rolls over). The gaseous mixture of fuel vapour and air
in the tank can exit through both the vent tube and the rollover valve during the filling process.
There are three major phases for the normal refueling process and these are indicated in
Figure 2, which is a graph of the tank gauge pressure versus time.
During the first phase of the filling process, the sudden decrease in the vapour space due to the
additional mass entering the vapour space causes the gases within the space to compress. There
is some evidence [2] that significant compression waves occur within the vent tube and rollover
valve tubing. This restricts the flow of gas through these tubes. The pressure in the vapour
space lowers as the exiting gas flow rate increases. During the second phase of a normal refilling
process, the tank gauge pressure remains almost constant. This phase constitutes most of the
refilling process. During the third phase of the refilling process, the liquid in the tank covers
the returning vent tube port. This decreases the venting area to only the rollover valve area.
This sudden decrease in venting area causes a sudden increase in the vapour space pressure. Due
to the high backpressure, the liquid in the filler tube backs up and the liquid in the vent tube
rises. Either situation will eventually lead to the shut-off sensing port in the fuel nozzle being
covered with liquid fuel, thus stopping the fuel flow. With the cessation of flow into the system,
Automotive Fuel Tank 521

the rollover valve quickly vents the excess air, reducing the pressure in the vapour space until it
reaches equilibrium with the surroundings.
Certain situations occur where the filling process is considered to be unsuccessful. Spill back is
one such situation. Spill back can be broken up into two categories: spit-back and well back [3].
Spit back occurs when small droplets of fuel exit the filler tube entrance with the escaping gases.
Well back occurs when slugs of liquid escape the filler tube entrance. Another unsuccessful filling
situation is premature shut-off. In this case, the fuel nozzle shuts off before the tank is completely
full.
In literature, there are few studies that consider the entire refueling process due to the difficul-
ties in modelling the process. Most attempts focused on one or more of the components in the
system. Rodriguez et al. (41 presented one- and three-dimensional models to predict vapour gen-
eration in a fuel tank due to diurnal, hot soak, running losses, and refueling losses. They used a
commercial CFD package (PHOENIX 1.6 version). Several articles have been published that deal
with vapour generation in general terms [5-7). Stoneman [3] used a commercial computational
fluid dynamic (CFD) package to simulate the filler pipe but gave little detail of the results. The
paper by Sinha et al. [2] also used CFD in order to describe the flow during refueling. Due to
the difficulty of conducting a CFD simulation of the entire system, Sinha et al. broke the system
up into several carefully selected key subprocesses. The work by Banerjee et al. [8] involved the
use of computational fluid dynamics to model the filler tube without vapour generation. They
found that due to the varying time and geometric time scales, modeling the entire system using
CFD z &ware was impractical. Banerjee et al. [9] also used CFD to model the filler t.ube with
vapour generation. Again, the model did not consider the tank. The pressure from the tank was
given a value and used as a boundary condition for the filler tube.
Several papers have been written concerning automotive fuel tanks that were not explicitly
about the filling process. They include, however, certain phenomenon involved in the fuel tank
filling process. One of these phenomena is vapour generation due to the evaporation of liquid
fuel. The model developed by Lavoie et al. [lo] predicts the amount of evaporative emissions.
The model includes the fuel tank, fuel pump, filler tube, liquid supply and return lines, fuel rail:

Table 2. Sensitivity analysis results.

Phase I Peak
522 S. FACKRELL et al.

vent valves, vent line, carbon canister, and purge lines. Lockhart [ll] presented three models for
vapour generation from a general tank. They include a model baaed on theoretical principles,
a semiempirical model and a model that was entirely empirical. The models were applied to
vapour emitted from a fuel tank and then compared to experimental data collected by Lockhart.
Kunimitsu et al. [12] provide methods of predicting vapour generation over a wide range of fuel
temperatures and compared them with available experimental test results. This model can be
used for predicting evaporative emissions and running losses. Parra et al. [13] used a three-
dimensional commercial CFD package to model the evaporative emissions from a fuel tank and
then compared the results with simple experiments. Their experimental work includes limited
data on the amount of gasoline mass evaporated.
Several papers have been published which deal with the general problem of tank filling (14,151,
however not related to automotive applications.
The work by Mastroianni [16] contained a set of experiments that indicated how certain pa-
rameters affect premature shut-off. For these experiments, a simple rectangular tank was fitted
with a standard filler pipe as well as a vent tube and a rollover valve. Mastroianni changed the
volume flow rate of the fuel, the RVP of the fuel, and the diameter of the vent tube. The tank
dome pressure and the filler tube pressure were measured using a pressure transducer.
The current work involves the development of a simplified lumped parameter model that in-
cluded the entire fuel system. The model also includes a vapour generation model presented by
Lavoie et al. [lo], a two-phase flow model presented by Wallis [17], and the idea of choking in
the vent tube that was inspired by the paper by Sinha et al. [2]. The experimental geometry
and conditions used by Mastroianni were taken as the basis for the simplified lumped parameter
model and the expe.rimental data of Mastroianni used for comparison.
The details of the fuel tank filling model are presented next, followed by a comparison of the
model prediction to selected experimental results of Mastroianni. The model is then used to
perform a sensitivity analysis of geometric as well as fluid mechanic parameters.

2. MODEL OF FUEL TANK FILLING


The mathematical model that is developed in this section is based on a simplified fuel tank
geometry as indicated in Figure 3. This model contains four control volumes and four valves.

Figure 3. Schematic diagram of the simple fuel tank system showing various control
volumes.
Automotive Fuel Tank 523

The flow throughout the system is considered to be incompressible unless otherwise stated. The
temperature of the system is considered to remain constant. Each control volume and connecting
tube arrangement will be considered individually as follows.

2.1. Control Volume 1


Control Volume 1 represents the annular area of the filler tube around the liquid that flows
from the nozzle that is approximated as a liquid column. This liquid column is assumed to be the
same diameter as the nozzle. The pressure in the annular area changes due to the flow of air/fuel
vapour through the opening around the nozzle, flow through the vent tube that is connected to
the tank, and flow into Control Volume 2. It is also affected by the movement of the interface
located between Control Volumes 1 and 2.
Applying the conservation of mass principle to Control Volume 1 and rearranging yields an
equation for the pressure in Control Volume 1 as,
RuTatm/MMair
” = s (A2 - AN) (hN - h2)
(1)

{[
(A2 - AN) ~1%
1 + [~voAvoV,o] + [pltA&t]

In this equation, Ru is the universal gas constant, Tatm is the atmospheric temperature, MM,i,
- [~&QL] } dt.

is the molecular mass of air, A2 is the cross-sectional area of the filler tube, AN is the cross-
sectional area of the nozzle, hi is the height of the nozzle from the bottom of the filler tube,
h2 is the height of the homogeneous mixture in the filler tube, pr is the density of the fluid in
Control Volume 1, pve is the density of the fluid flowing through the space around the nozzle
at the opening of the filler tube, A ,,e is the cross-sectional area of the space around the nozzle
at the opening of the filler tube which is assumed to be constant, V,O is the velocity of the fluid
through Avo, p4 is the density of the fluid flowing through the vent tube from the tank which is
assumed to go only from the tank to the filler tube, Avt is the cross-sectional area of the tube
connecting the tank to the filler tube which is assumed to be constant, VT,,,is the velocity of the
fluid through the vent tube that is attached to the filler tube, (Y is the void fraction, and QL is
the volume flow rate of liquid fuel from the dispensing nozzle.

2.1.1. Velocity through Valve 0


Applying the modified Bernoulli’s equation across the opening between the dispensing nozzle
and the filler tube results in the following equation for the velocity through the opening. This
equation takes into account the wall friction, the entrance loss, and the exit loss.

(2)

Paem is atmospheric pressure, fvc is the friction factor for Valve 0 (see the section for Valve l),
L,,c is the effective length of Valve 0, and D,,o is the diameter of Valve 0.

2.2. Control Volume 2


As the pressure in the tank rises, the flow of fluid out of the filler tube slows while the flow
into the filler tube remains constant. This causes the fluid to collect in the filler tube. This
collection of fluid in the filler tube is contained in Control Volume 2. It has been assumed the
cross-sectional area of the filler tube is constant. It is also assumed that the two-phase flow in
the filler tube is homogeneous and incompressible. Applying the conservation of mass principle
to Control Volume 2 and rearranging yields

h2 = i /- (Q1-2 - Q2-3) dt. (3)


524 S. FACKRELL et al.

In this equation, Qi--2 is the volume flow rate of air/fuel vapour and liquid flowing into Control
Volume 2 and Q2-3 is the volume flow rate of homogeneous mixture flowing from the filler tube
into the tank.

2.2.1. Valve 1

The losses for the flow in the filler tube are accounted for through the use of Valve 1. Applying
the modified Bernoulli’s equation and rearranging yields the following equations for the volume
flow rate. If the height of the liquid in the tank is lower than the entrance to the filler pipe, the
equation is

Q2-3={[y+h2] [ ’ ]; (4)
(.fvlLl/~u1+ Kexit + Kent) (1/2gAz) - 1/2gA$

where, P4 is the pressure in the tank dome and oh is the specific weight of the homogeneous fluid.
If the height of the liquid in the tank is greater than the entrance to the filler pipe, the modified
Bernoulli’s equation yields

Q2-3=
9 -P4%

%
(h3 - hf) + h2

I[ (fvlL~l/Dvl + Kexit + ke,) (1/2gAi) - 1/2gAg 1 ’ (5)

2.3. Control Volume 3


After the fluid reaches point e (see Figure 3), it is assumed that the liquid and gases separate
with liquid falling to the bottom of the tank and the gases rising to the top of the tank. Control
Volume 3 contains the collection of the liquid fuel in the tank. A small amount of fuel in the
tank is assumed to initially be in the tank.
Applying the conservation of mass to Control Volume 3 and rearranging yields

h3 = i Q2-3 - hi- “RrT Aliq (P4,v-eq - P4,v) dt.

In this equation, ha is the height of the liquid in the tank, A3 is the cross-sectional area of the
tank, h,j is the mass transfer coefficient [7], MM,, is the molecular weight of the fuel vapour [7],
Ali, is the surface area of the liquid gas interface, P4,v--eq is the equilibrium vapour pressure for
the fuel vapour [7], and P+ is the partial pressure of the vapour in the tank dome.

2.4. Control Volume 4


Control Volume 4 represents the air and fuel vapour mixture that is collected in the tank.
Applying the law of partial pressures to Control Volume 4 for air and fuel vapour gives

P4 = P4,v + P4,a,

where P4 is the pressure in the tank dome (Control Volume 4) and P4+ is the partial pressure of
the air in the tank dome.
There are two sources of mass addition to the tank vapour space; the air/fuel vapour that
enters through the filler tube, and the fuel vapour that evaporates from the liquid fuel. There are
two exit ports for the gases; the vent tube which is connected to the filler tube and the rollover
valve. As air and fuel vapour collect in the tank, the tank pressure rises, causing the liquid in
the filler tube to backup.
Applying the conservation of mass of air to Control Volume 4 yields

p = RuT’tm/MMair
4,a ~4.oA4$f-(l-Z4) ( 62,out +%3,out) + (I- 2,) aplQ2-3 4 (8)
I A4 (H-h)
Automotive Fuel Tank 525

where Ad is the cross-sectional area of the tank, H is the height of the tank, p+, is the density of
air in the tank dome, 2, is the mass fraction of fuel vapour in the vapour space, ~u3,01~~tiv2,~u~
is the mass flow rate through the vent tube and is the mass flow rate through the rollover valve.
The mass fraction of fuel vapour in Control Volume 1, 21, is given by,

Applying the conservation of mass of fuel vapour to Control Volume 4 yields

- P4,v) + p4,v44%
p4,v = (W
s
-24 (+12,out + 7%3,0~t) + Z1(TYplQ2-3
where ~4,~ is the density of fuel vapour in the tank dome.

2.4.1. Mass fraction


In order to determine the mass fraction of fuel vapour in the vapour space, Gibbs-Dalton law
for perfect gas mixtures [18] and the ideal gas law may be combined to give

1
z4 = -MM,,,
MMipP4,v
[ P4,uMMvap
+ (P4 - P4,v) MM:,,
+ (P4 - P4,v) M&r 1
P4,v
P4’
(11)
2.4.2. Equilibrium vapour pressure
From the article by Lavoie et al. [7], the equilibrium vapour pressure can be expressed as

ln(E!) (+
P~+~~(RVP, T, Z to v) = pfocexp 1 (12)
%vP+A(ZI to*-&VP)

In this equation, Pfo, is the focal pressure, Z’rO,is the focal temperature, Tuvp is the temperature
that RVP is measured at, T is the temperature of the system and 2~ to 21is the mass fraction of
liquid that evaporates to vapour given by
1 1
z ptRuT (13)
1 to v = 1+ 7vIL);;~~MM =1f( W/h3)-lU’4,v,,MMva, ’
VW
and &vp is the mass fraction for the conditions used to determine RVP,
1
z RVP = 1 + pLRuT ’ 114)
I.RVP.MM,,,

2.4.3. ‘ielve 2
The vent tube and the vent tube valve were arranged as indicated in Figure 4. Point f is a
point in the tank dome, Point v2 is just inside the vent tube valve, point vt is a point just inside
the vent tube at the connection to the filler tube and point h is a point inside the filler tube
(Control Volume 1)
It is assumed that the losses for the vent tube are accounted for in Valve 2.
The modified Bernoulli’s equation between point f (the tank dome) and h (the filler tube) (see
Figure 4) may be written as

Q4-1
=11, (15)
526 S. FACKRELL et al.

tl.
C.V.
0

Figure 4. Diagram of the vent tube.

2.4.4. Valve 3

It is assumed that the losses for the rollover valve are accounted for in Valve 3.
The modified Bernoulli’s equation between Point f and g (see Figure 4) is given by,

(16)

2.4.5. Choking

The flow of air/fuel vapour mixture that passes through the rollover valve and vent tube has
been determined in previous sections assuming the flow to be incompressible. In this section, the
conditions required for the flows in the vent tube and rollover valve to experience choking are
estimated based on a pseudo-compressible flow analysis. The article by Sinha et al. [Z] describes
the effect of small droplets in the vapour, which has the effect of lowering the speed of sound.
With a lower speed of sound the occurrence of choking is possible. To be rigorous, the effects
of density changes throughout the flow should be taken into account. In this case, however, the
choking condition is approximated utilizing the velocity calculated with the incompressible flow
model and comparing it with a constant speed of sound. When the velocity is equal to the speed
of sound, choking is said to occur. It is possible that flow leaving the vapour space via the vent
tube and the rollover valve to experience choking.
According to Wallis [17], if the density of the gas phase, ps, is much less than the density of
the liquid phase, ,OL, and the speed of sound in the gas, cs, is less than the speed of sound in the
liquid, CL, the speed of sound in the mixture can be given by

=d +%
Gnix -Y)' PLY(1
(17)

In this equation, Y is the percentage by mass of liquid droplets in the gas. The lowest value of
c,ix occurs at Y = 0.5 and is given by

The minimum speed of sound in the air vapour mixture will, therefore, be 26.85m/s.
Automotive Fuel Tank 527

If the velocity of the flow through the rollover valve is equal to the speed of sound in the air
vapour mixture (Vu3 = cmix), the mass flow rate through the rollover valve can be obtained as
follows:
%J3,out = ~$&0.097644 = ~A~0.097644. (19)
m,x

If the velocity of the flow through the vent tube (Valve 2) is equal to the speed of sound in the
air vapour mixture (V,z = cmix), th e mass flow rate in the vent tube becomes

ti,2,0ut = dA&0.097644 = dAg+0.097644. (20)


m,x

3. SOLUTION OF MODEL EQUATIONS


The model presented in Section 2 can be reduced to five integral equations. These equations
are solved, simultaneously, for the pressure in the filler tube, 9, the partial pressure of air in the
fuel tank vapour space, P+, the partial pressure of vapour in the fuel tank vapour space, Pd,,,,
the height of the liquid in the filler tube, ha, and the height of the liquid in the tank, h3. The
following initial conditions were used; &Patm, P4+,=Patm/2, P+=Patm/2, h2 = 0, and h3 = 0.
The thesis by Fackrell [19] contains the complete details of the solution.
In order to solve the set of ordinary differential equations, a numerical procedure was employed.
A feature of Version 5.3.1.29215a (R1l.l) of the MATLAB program, referred to as Simulink, was
used. It was installed on an IBM compatible PC with an 800MHZ AMD Athlon processor and
640MB of RAM memory. Since this model is considered stiff, the implicit solver “ode23s” was
used. The “ode23s” solver is a modified implicit Itosenbrock method of orders 2 and 3 with error
control for stiff systems [20]. This is a variable step size solver with the maximum step size set
to 0.01 seconds and initial step size set to 0.001 seconds.

4. APPLICATION OF MODEL
The model was used to predict the filling of the fuel tank conditions that were experimentally
studied by Mastroianni [16]. The same geometry, flow conditions and fluid properties were mod-
elled. This involved a tank with a cross-sectional area of 0.2787m2 and a height of 0.305m as
well as a filler tube that was 0.5m long with a diameter of 0.0318m. This model also included
the provision to change the vent tube diameter, fuel RVP value and the fuel dispensing nozzle
volume flow rate and a number of fluid properties. It is important to note that certain param-
eters could not be accurately determined. For example, the mass transfer coefficient could only
be approximated due to the multiphase, multicomponent flow that is a highly turbulent flow.
Likewise, the equivalent diameter of the vent area between the fuel dispensing nozzle and the
filler tube had to be estimated.
The flow from the dispensing nozzle was simulated as a step function both for the turn-on and
shut-off. The height of the liquid in the filler tube, hz, was used as the parameter to determine
the shut-off condition. Once h2 became equal to the height of the filler tube (0.5m), the fuel
from the dispensing nozzle was shut-off.
Although the thesis of Fackrell[19] contains a comparison with all of the experiments conducted
by Mastroianni [16], space limitations only allow selected examples to be presented here. It has
been decided to select, for presentation, two examples of each of the distinctly different filling
conditions; normal shut-off (NSO) and premature shut-off (PSO). In each of these cases, an
example that represents the best comparison and one that represents the worst comparison with
experimental data are included. The tank dome pressure was used as the basis for comparison
as it was experimentally available. The pertinent model parameters that were used for these
examples are listed in Table 1.
528 S. FACKRELL et al.

The current model contains a number of parameters, some of which affect the results more
than others. A sensitivity analysis was conducted to determine the effect that the parameters
have on the results. The parameters which were investigated are the fuel volatility as indicated
by the RVP, the liquid fuel dispensing flow rate, vent tube diameter, effective diameter of the
vent area around the fuel nozzle, filler tube diameter, rollover valve diameter, void fraction, and
the mass transfer coefficient. The simulation was run with nominal values of the parameters
(Dvz = 6.4mm, Qin~ = 45L/min, RVP = 55kPa), then at values of *lo%, &l%, or &O.l%,
depending on the sensitivity of the parameter.

Table 1. Parameters used in example cases.

Vent Tube Diameter Volume Flow Rate


Example Case
(mm) (L/min)

NSO #l 9.5 45 72
NSO#2 9.5 38 57
PSO #l 3.2 38 83
PSO #2 3.2 45 82

The effect of including the assumption of a “pseudo-choking condition” in the vent tube and the
rollover valve was also investigated. The simulation was run with nominal values of the parameters
(D,,z = 6.4mm, Qin L = 45 L/min, RVP = 55 kPa). Four conditions were investigated. The
first case considered choking in both the vent tube and the rollover valve. The second case only
included the effects in the vent tube valve. The third case considered the effects of choking in only
the rollover valve. The fourth case neglected choking effects completely. The simulation was also
run with nominal values of the parameters (D,,z = 3.2 mm, &in L = 45 L/min, RVP = 55 kPa)
with and without choking in the vent tube and valves. These results are included in the next
section.

5. RESULTS
Figures 5 and 6 show the results of a refueling process with normal shut-off. Figure 5 is an
example of a good agreement between the experimental results and those of the lumped parameter
model. The peak value of pressure during Phase I as well as its time of occurrence and the value of
the relatively constant Phase II pressure are modelled accurately. The time taken to completely
fill the tank is slightly in error, which could very well be due to the estimate of value of the height
of the fuel cut off sensor height.

5000
4500
4000
3500

0 20 40 60 60 100 120

Time (8)
.

--t Model - Experimental

Figure 5. Tank dome pressure, D,2 = 9.5 mm, Qi,, L = 45 L/min, RVP = 72 k Pa.
Automotive Fuel Tank 529

a- Model +- Experimental

Figure 6. Tank dome pressure, D,,2 = 9.5 mm, Qin L, = 38 L/min, RVP = 57 k Pa

Figure 6 is an example of the poorest agreement between the experiment and the model pre-
diction. The magnitude of the peak pressure in Phase I is seen to exhibit a considerable amount
of error, while the time at which the peak occurs is predicted to a reasonable degree of accuracy.
The magnitude of the steady pressure in Phase II is reasonably well predicted and the time taken
to completely fill the tank is in approximately the same amount of error as in the previous case.

0 2 4 6 8 10 12
Time (s)

= Model +- Eperimental

Figure 7. Tank dome pressure, D,Q = 3.2 mm, Qin L = 38 L/min, RVP = 83 k Pa.

8000
f 7000
f 6000
t 5000
P 4000
t3000
8 2000
r 1000
c
0 3

4 6 8 10 12 14

Tifns (s)

a- Model -+ Experimental

Figure 8. Tank dome pressure, D,2 = 3.2 mm, Qin L = 45 L/min, RVP = 82 k Pa.
530 S. FACKRELL et al.

Table 2. Sensitivity analysis results.

Parameter Phase I Peak

of Vent Area Around

Fuel Volatility

Coefficient
Void Fraction f10 730 M0 X0
Filler Tube
f10 ZO M0 X0
Diameter

Figures 7 and 8 include comparisons of the mathematical model and the experimental results
for the case of premature shut-off in the best and worst cases, respectively.
In Figure 7, the variation of pressure with time is predicted to a reasonable degree of accuracy.
This is also true with the occurrence of premature shut-off and the time at which it occurred.
The variation of pressure after the shut-off occurs is not predicted in the model and, hence, no
comparison should be made. It is interesting to note that the tank dome pressure is actually
decreasing when shut-off occurs due to a phase lag between the tank dome pressure and with the
filler tube level.
Figure 8 displays an example of a poor agreement between the lumped parameter model and
experimental results. The predicted tank dome pressure closely follows the experimental values
until premature shut-off occurs. The model underpredicts the time at which this occurs by
approximately 2.5 seconds. The small times that are involved in premature shut-off, however,
result in a large percentage error in the prediction of this quantity. The lower value of shut-off
time results in a lower value for the peak pressure value.
Although the prediction of the tank dome pressure ranged from good to only fair, the occurrence
of PSO or NSO was correctly predicted in every case. This was not only true in the cases presented
in this paper but for each case considered by Mastroianni [16].
The results of the sensitivity analysis are presented in Table 2. From this table, it is evident
that the tank dome pressure results were found to be very sensitive to changes in the effective
diameter of the vent area around the fuel nozzle, slightly less sensitive to the liquid fuel volume
flow rate, and fuel volatility. The results were only moderately sensitive to the rollover valve
diameter. The results, however, were not sensitive to changes in the vent tube diameter, mass
transfer coefficient, void fraction, and filler tube diameter. The insensitivity to the mass transfer
coefficient is particularly important since this was the parameter that was least well known.
The results of the cases with different conditions for choking are presented in Figure 9. Choking
in the rollover valve has the effect of increasing the maximum Phase I peak pressure that can
lead to premature shut-off. Choking in the rollover valve, however, does not affect the tank dome
pressure during Phase II of filling or the normal shut-off time.
Automotive Fuel Tank 531

f
0 20 40 TiniO(r, 60 100 120

+Nominal a- Without Choking


+Wiihout Choking in Dd2 +Without Choking in Dv3

Figure 9. Tank dome pressure with/without choking for normal shut-off.


5000 ‘I

0 20 40 60 60 100 120
Time (8)

--c Choking + Non-Choking

Figure 10. Tank dome pressure with/without choking for premature shut-off.

The variation of tank dome pressure with time for the case of &in L = 45 L/min, RVP = 55 k Pa
but with a 3.2mm diameter vent tube for the cases of choking (in both vent tube and rollover
valve) and no choking are included in Figure 10.
In this case, the inclusion of choking effects makes the difference between predicting premature
shut-off or not. Premature shut-off was also observed experimentally for this condition.

6. CONCLUSIONS
A simplified lumped parameter numerical model that reasonably describes the process of refu-
eling an automotive fuel tank in order to predict premature shut-off has been developed. Unlike
previous work, this model includes the entire fuel system and yields results quickly. A comparison
of the model results with available experimental data indicated a good to only fair agreement,
although the model predicted premature shut-off in every experimental case considered.
The sensitivity of the results to changes in the problem parameters indicated that:
l the results were very sensitive to changes in the effective diameter of the vent area around
the fuel nozzle, liquid fuel volume flow rate, and fuel volatility;
l the results were sensitive to the rollover valve diameter;
l the results were not sensitive to changes in the vent tube diameter, mass transfer coeffi-
cient, void fraction, and filler tube diameter;
l including the effect of choking has a significant affect on the prediction of Phase I peak
pressure and is important for accurately predicting the nature of the shut-off.
532 S. FACKRELL et al.

Inclusion of the effect of choking in the vent tube and rollover valve was shown to have a
significant affect on the prediction of Phase I peak pressure and is important for accurately
predicting the nature of the shut-off.

REFERENCES
1. M. Holloway, Fill ‘er up, Scientific American, 92-93, (May 2000).
2. N. Sinha, R. Thompson and M. Harrigan, Computational simulation of fuel shut-off during refueling, SAE
Technical Paper Series, Paper 981377, Society of Automotive Engineers, (1998).
3. S. Stoneman, On the design of automotive fuel filler pipes, Automotive Engineer 22 (l), 32-36, (1997).
4. M.A. Rodriguez, M.T. Parra and F. Castro, Vehicle evaporative emission models, In ECCEMEI 1994 Pro-
ceedings, (1994).
5. S.R. Reddy, A new correlation for predicting average vehicle evaporative emissions as a function of fuel
volatility, SAE Technical Paper Series, Paper 881592, Society of Automotive Engineers, (1988).
6. M.F. Bardon and V.K. Rao, Estimating instantaneous properties of vaporizing light petroleum fractions,
Combustion and Flame 84, 427-431, (1991).
7. G.A. Lavoie and C.S. Smith, Vapor pressure equations for characterizing automotive fuel behavior under hot
fuel handling conditions, Society of Automotive Engineers, (1997).
8. R. Banerjee, K.M. Isaac, L. Oliver and W. Breig, A numerical study off automotive gas tank filler pipe two
phase flow, SAE Technical Paper Series, Paper 2001-01-0732, Society of Automotive Engineers, (2001).
9. R. Banerjee, X. Bai, D. Pugh, K.M. Issac, D. Klein, J. Edson, W. Breig and L. Oliver, CFD simulation
of critical components in fuel filling systems, SAE Technical Paper Series Paper 2002-01-0573, Society of
Automotive Engineers, (2002).
10. G.A. Lavoie, Y.A. Imai and P.J. Johnson, A fuel vapour model (FVSMOD) for evaporative emissions system
design and analysis, SAE Technical Paper Series, Paper 982644, Society of Automotive Engineers, (1998).
11. M.C. Lockhart, Predicting tank vapor mass for on-board refueling vapor recovery, SAE Technical Paper
Series, Paper 970308, Society of Automotive Engineers, (1997).
12. M. Kunimitsu, K. Mori and T. Muragushi, Prediction of gasoline vapor generation from a vehicle fuel tank
at high temperature, JSAE Review 14 (4), 6548, (1993).
13. M.T. Parra, M.A. Rodriquez, F. Castro and J.M. Villafruela, Three-Dimensional Evaporative Emission
Model, IMechE, C496/005/95, (1995).
14. D.A. Vaughan and G.R. Schmidt, Analytical modeling of no-vent fill process, Journal of Spacecraft and
Rockets 28 (5), 574-579, (1991).
15. P.B. Whalley, Two-phase flow during filling and emptying of bottles, International Journal of Multi-Phase
Flow 17 (l), 145-152, (1991).
16. M. Mastroianni, Experimental investigation of automotive fuel tank filling, M.A.Sc. Thesis, Mechanical,
Automotive and Material Engineering Department, University of Windsor, Windsor, Ontario, (2000).
17. G.B. Wallis, One-Dimensional Two-Phase Flow, McGraw-Hill, New York, NY, (1969).
18. H.A. Hasanein, M.S. Kazimi and M.W. Golay, Forced convection in-tube steam condensation in the presence
of non-condensible gasses, International Journal of Heat and Mass Transfer 39 (13), 2625-2639, (1996).
19. S. Fackrell, A lumped parameter model for the filling of an automotive fuel tank, M.A.Sc. Thesis, Mechanical,
Automotive and Material Engineering Department, University of Windsor, Windsor, Ontario, (2001).
20. R. Ashino, M. Ngase and R. Baillancourt, Behind and beyond the MATLAB ODE suite, Computers Math.
Applic. 40 (4/5), 491-512, (2000).

You might also like