Você está na página 1de 11

ILASS Europe 2011, 24th European Conference on Liquid Atomization and Spray Systems, Estoril, Portugal, September 2011

Comparison of current Spray models under high pressure and high temperature engine
relevant conditions
T. Vogel
*12
, S. Rie
12
, M. Lutz
23
, M. Wensing
12
, A. Leipertz
123

1: Institute of Engineering Thermodynamics, FAU Erlangen-Nrnberg, Germany
2: Erlangen Graduate School of Advanced Optical Technologies, Germany
3: ESYTEC Energie- und Systemtechnik GmbH, Erlangen, Germany
Abstract
The liquid spray length is a key parameter in the optimization of diesel combustion processes. Since 1960
several spray models have been developed in order to describe the spray propagation with and without the influ-
ence of the temperature of the ambient gas atmosphere. However, since the time most propagation models were
developed injection systems and the thermodynamic conditions under which the fuels are injected have changed
much.
The present investigation compares more than ten spray models to measurements performed in a high tem-
perature and high pressure cell in order to simulate the conditions in modern diesel engines using two different
modern injectors.
In addition to the penetration-model-comparison the nozzle exit velocity is determined via the differentiation
of an origin-shifted root of time curve fitted to measured data. The results are compared to the corresponding
Bernoulli-velocity in order to calculate the discharge coefficient. The discharge coefficient is compared to the
pressure drop of the nozzle and the discharge Reynolds number.
In order to simulate different engine operation different ambient conditions were defined into which the
spray was injected. The effect of each single parameter on the spray formation can be identified since the pa-
rameters are controlled separately. These parameters varied are the ambient gas temperature (523 K, 623 K,
723 K, 823 K and 923 K), the ambient gas pressure (1 MPa, 3 MPa, 5 MPa, 7 MPa and 9 MPa) and the fuel
pressure (40 MPa, 80 MPa, 120 MPa, 160 MPa and 200 MPa) while fuel temperature (363 K) and injection
duration (450 s) were kept constant. The injection chamber was permanently scavenged with nitrogen thus
ignition and combustion were suppressed, even at high temperatures. Two injectors from two different vendors
were used. The first injector is a piezo-actuated servo-hydraulic 8-hole injector (Bosch) with orifice diameter of
152 m while the length of the nozzle is 1 mm. The second injector is a 7-hole and also piezo-actuated and
servo-hydraulic injector (Continental). The orifice diameter of this injector is 120 m and the length of the noz-
zle is 720 m.
In order to measure the spray parameters like penetration and spray cone angle a Mie-scattering setup con-
sisting of four flash lamps and one sensitive CCD-camera was installed. At every operating condition a total of
16 points of time after SOI and 32 images for each point of time were recorded.
Firstly, spray models from literature were compared to identify main parameters influencing the penetration
depth vs. time. A dimension analysis is given in Table 1. The main parameters are fuel pressure difference, ori-
fice diameter and gas density. The exponent of time given varies from 0.48 to 0.64. Some models include also
the orifice length and the gas temperature.
All operating points were compared to all spray models. Under the present conditions the models of Taylor
and Walsham, Dent, Lustgarten, Hiroyasu and Arai, Desantes, Arregle and Sazhin were much closer to the
measurements data than the rest of the models, especially at low temperature conditions. At high temperature
and pressure conditions all models differ much from the measurements. The measured penetration depth reaches
a stationary state, where injection rate and evaporation rate reach the same value so that the penetration length
stays nearly constant. None of the models includes the effect of evaporation completely. All models predicted a
continuously increasing penetration depth. Overall, no model predicted the spray accurate at all operating condi-
tions although some models are close to the measured data at some operating points. The determination of the
nozzle exit velocity allowed the calculation of typical, operating point dependent, discharge coefficients. Espe-
cially under high injection pressure conditions the analysis carried out a significant difference in the throttling
behavior which is not only caused by the orifice diameter and length but also by the injection system itself.
Introduction
In modern internal combustion engines the fuel is usually injected directly into the combustion chamber. In
contrast to an external mixture preparation the requirements for the injection systems are more complex. The
injection system directly affects the local fuel/air ratio, the ignition and combustion behavior and therefore the

*
Corresponding author: thomas.vogel@ltt.uni-erlangen.de
1
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
power generation, fuel consumption and emission formation, while multiple injection strategies dramatically
increase the number of operating parameters.
The nozzle exit velocity is a parameter of primary importance. It affects not only directly the (primary and
secondary) atomization but also the mixture preparation due to the momentum brought into the combustion
chamber. Several models have been developed in order to describe the penetration depth vs. time in relation to
the thermodynamic conditions like fuel pressure, gas pressure, gas temperature and nozzle geometry. The first
models were developed in the 1960s [1, 2]. However, the conditions existing in todays internal combustion
engines, especially in DI compression ignition engines, are dramatically increased compared to the conditions
under which these models were emerged.
Existing spray models
A short overview of the existing spray models is presented in chronological order.
In 1960 Wakuri et al. [1] developed a semi-empirical/semi-analytical model based upon a measurement se-
ries where diesel fuel was injected with pressures up 76 MPa into an gas atmosphere up to 873 K and 2.4 MPa.
(1)
In 1964 Sitkei [2] determined that single droplets behave different than a spray jet. His equation is the only
one to consider the viscosity of the fuel covering thermodynamic conditions up to gas pressures at 4.8 MPa and
temperatures up to 873 K. The fuel was injected with up to 58.8 MPa.
(2)
Taylor and Walsham [3] made the next attempt and developed a model which was adjusted over the years. It
is the only model including the influences of the nozzle diameter/length ratio.
(3)
Dent published his model in 1971 [4]. He was the only one considering the gas temperature. His original
model was published with the Rankine temperature scale. In this paper the adapted model will be used.
(4)
A next semi-empirical model was defined by Lustgarten [5] in 1973 based upon a gas jet theory and an
equivalent nozzle diameter.
(5)
Hiroyasu and Arai [6] published their model in 1990. The model divided the spray propagation in a nozzle
near- and a nozzle far region. The border in between these two is the breakup time. Due to the high injection
pressures the break up time is very short and therefore negligible in this comparison.
(6)
2
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
Naber und Siebers [7] were the first to develop a spray based upon a physical model. Just like the model
from Hiroyasu and Arai the model is divided in a near nozzle- and a far nozzle region. And also here the break
up time is very short due to todays high injection pressures. So here only the nozzle far model is considered.
(7)
Desantes [8] also investigated the influence of viscosity and surface tension on spray behavior. Therefore he
used fuels with different thermo physical parameters. However, it became clear that these parameters have no
influence on the spray propagation.
(8)
Arrgle [9] used a dimension analysis and the -Buckingham theorem to develop a model for spray penetra-
tion. He didnt use units and therefore the function gives only proportional behaviour. In order to use this model
a factor has been calculated out of a single measurement with 1.927.
(9)
Sazhin [10] set his focus on the base physics. He also devided the spray propagation in two stages, the first
stage where the drag on the droplets dominates and later the two phase flow. He also came to the conclusion that
under todays conditions the first stage is negliable. For the two phase flow he published the following equations.
The first equation gives the detailed model and the second a simplified version. He also considered the initial
velocity in dependence of the discharge coefficient of the nozzle. In order to test his equation, he used different
discharge coefficients from literature [11-13] and also determined one for the Newtonian flow.
(10)
(11)
(12)
(13)
In Table 1 a dimension analysis is given for all spray models. It becomes clear that for most exponents all
models give similar values. This can be especially seen for the influence of time after the start of injection. Here
the values range only from 0.48 to 0.64 besides the values that found Hiroyasu, Arai, Naber and Siebers for the
3
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
initial stage. In none of the models typical fluid parameters like viscosity and surface tension were found to have
an effect on the spray formation vs. time.
Table 1. Existing Spray models
time d
0
l
0
/d
0
gas density
gas
temperature
injection
pressure
influence parameter
(exponent n)
t
n
(d
0
)
n
(l
0
/d
0
)
n
(
G
)
n
(T
G
)
n
(p)
n
Wakuri et al. (1960) [1] 0.50 0.50 X -0.25 X 0.25
Sitkei (1964) [2] 0.48 0.82 X -0.35 X 0.53
Taylor and Walsham (1969) [3] 0.64 0.36 0.18 -0.32 X 0.32
Dent (1971) [4] 0.50 0.50 X -0.25 -0.25 0.25
Lustgarten (1973) [5] 0.54 0.46 X -0.23 X 0.27
Hiroyasu and Arai (1990) [6] 1; 0.50 X; 0.50 X X; -0.25 X 0.50; 0.25
Naber and Siebers (1996) [7] 1; 0.50 X; 0.50 X X; -0.25 X 0.50; 0.25
Desantes et al. (1998) [8] 0.50 0.50 X -0.25 X 0.25
Arrgle et al. (1999)[9] 0.57 0.37 X -0.41 X 0.26
Sazhin et al. (2001) [10] 0.50 0.50 X -0.25 X 0.25
Materials and Methods
Measurement setup
The present measurement series, like former measurement series [14, 15], has been performed on the high
temperature high pressure optical combustion test rig (OptiVeP). This test rig is full optical accessible, perma-
nently scavenged, combustion chamber which can be run either with normal air, pure nitrogen or mixture of
these. Gas temperatures are possible up to 1000 K and pressures up to 10 MPa can be applied, while the injection
pressure can be adjusted up to 250 MPa.

Figure 1: Test rig of the optical injection chamber
Injectors and experiment program
All fuel penetration models include the nozzle diameter. To investigate this influence two injectors from two
different vendors with different nozzle geometry but with similar piezo servo-hydraulic principle have been
chosen.
4
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions

Figure 2: Evaluated Injectors Figure 3: Measurement matrix
The first injector, in the following text named Injector A, was manufactured by Bosch (Figure 2, right
side). It is the CRI 3.0 from a BMW 3-liter-6-cylinder-engine from 2006. Unlike to the engine, where the injec-
tion pressure is at 160 MPa, the injection pressure was set up to 200 MPa on the test rig. The second injector was
built by Continental and is named Injector B (Figure 2, left side). This injector is normally mounted in a 2-
liter-4-cylinder-engine from VW and is operated with a maximum injection pressure of 200 MPa. Both injectors
were tested at the same operating point and with the same injection timing. The injection timing was not adjusted
to the same fuel quantity. This was not necessary because these injectors were not tested on a real engine and
therefore the focus is set on the spray propagation vs, a defined period of time. The nozzle geometry differences
can be seen in Error! Not a valid bookmark self-reference..
Table 2: Nozzle geometry overview
Type d
0
[mm] l
0
[mm] # of holes [-]
A Bosch CRI 3.0 0.152 1.000 8
B Continental PCR 2 0.120 0.720 7

In Figure 3 the measurement conditions, gas temperature vs. gas pressure, can be seen. All operating points
have been defined thus only one parameter changes from one operating point to another. With this information it
is possible to identify the influence, either linear ore exponential, of each parameter. In addition to the conditions
given in this tabular, the fuel pressure has been varied in steps of 40 MPa up to 200 MPa. The injector tempera-
ture has been kept constant at all operating points at 373 K simulating a typical engine temperature. At all oper-
ating points images were acquired in a window of 1 ms, every 100 s after the visible start of injection (vSOI,
the point were the first spray to exit the nozzle can be seen). In order to get a higher resolution in the propagation
phase, images were taken every 25 s up to 200 s after vSOI. The number of images was set to 32 which gives
good information about the statistical behavior.
Evaluation procedure
In Figure 4 the evaluation procedure for the spray images is shown exemplary. The raw images are turned
with the #1 spray cone in 12 oclock position and a threshold is applied generating a binary spray image. Out if
this binary spray information all parameters, like penetration depth and spray cone angle, are measured. From all
32 images a mean image, an outline plot and frequency scale image are generated.

raw image mean image with
threshold applied
outline plot frequency scale image
Figure 4: Image processing procedure
The nozzle exit velocity itself is determined numerically out of the penetration depth vs. time data measured.
Therefore, a Matlab script has been written importing and processing the results of the spray measurement. Six
5
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
of the investigated spray models [1, 4, 6-8, 10] set the exponent for the time dependence at 0.5 which was also
chosen here. In earlier investigations [15] it was proven that the spray propagation can be described by a origin
shifted root of time function. The deviation of this function gives the spray propagation velocity vs. time and
thus the nozzle exit velocity at the initial point. The algorithm designed is able to separate the propagation phase
from the stationary spray state, where the penetration is nearly constant. To do this, the standard deviation of the
penetration depth is investigated. By entering the stationary stage, the standard deviation increases significantly.
Some injectors tend to show a throttled opening behavior, which leads to transient changes in the discharge
coefficient. In such cases a left bended penetration vs. time graph is visible in the beginning. These values are
not included in the function fitting algorithm. An exemplary result graph is given in Figure 5 with the original
graph (blue), the beginning of the stationary state (vertical black line) and the fitted function (green).


Figure 5: Result graph of the Matlab script (german)
6
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions

The discharge coefficient is calculated out of the theoretical Bernoulli velocity and the measured velocity:
(14)
with:
(15)
Results and Discussion
Spray model comparison
The spray model comparison clarified that the models from Wakuri, Naber and Siebers and Desantes deliv-
ered results differing factor two from the measured results and thus these models are not considered in the fol-
lowing results. In non-evaporating conditions nearly all models follow the measured results nearly exactly. Un-
der evaporating conditions, like in Figure 6 and Figure 7 the results are different. Until 200 s after vSOI for
injector A and 300 s after vSOI for injector B the models follow the measured graph within a certain tolerance.
Afterwards the measurements enter the stationary state while the models still assume a propagating spray.

Figure 6: Injector A (p
G
= 9 MPa/ T
G
= 723 K/ p
F
= 200 MPa/ T
F
= 363 K)
7
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions

Figure 7: Injector B (p
G
= 9 MPa/ T
G
= 723 K/ p
F
= 200 MPa/ T
F
= 363 K)
To compare the spray models, the sum of error squares (SoES) has been calculated and is given in Table 3. It
becomes clear that the deviation of the models differs with the injector used. While for injector A the Sazhin
equation with a C
d
of 0.44 gives the best fit, the model of Dent is the most accurate for the other injector. If other
thermodynamic conditions are applied, the results of this quality check differ unpredictable. Also none of the
models tested considers the effects of evaporation which is the main reason for the high error values.
Table 3: Sum of error squares of all evaluated models
model SoES Inj. A [mm] SoES Inj. B [mm]
Taylor and Walsham 19,92 30,91
Dent 39,08 16,13
Lustgarten 161,18 91,06
Hiroyasu and Arai 173,09 84,22
Arrgle 51,47 22,98
Sazhin et al. equation (12) C
d
= 0.7 124,31 16,84
Sazhin et al. equation (12) C
d
= 0.39 20,64 168,91
Sazhin et al. equation (12) C
d
= 0.44 9,42 113,33
Sazhin et al. equation (11) C
d
= 0.44 25,83 168,56
Initial spray velocity determination
Using the Matlab script delivers the nozzle exit velocity for injector A in Figure 8 and injector B in Figure 9.
Figure 8: Nozzle exit velocity Inj. A for diff. p
G
Figure 9: Nozzle exit velocity Inj. B for diff. p
G
Like expected the nozzle exit velocity increases with an elevation of the injection pressure, or pressure dif-
ference. While the injector A shows linear increase of the exit velocity and therefore, a nearly constant discharge
8
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
coefficients vs. injections pressure (Figure 10), the injector B shows an increase in the discharge coefficient at an
injection pressure of 120 MPa (Figure 11). This indicates a change in the nozzle/injector internal flow.
Figure 10: discharge coefficient Inj. A for diff. p
G
Figure 11: discharge coefficient Inj. A for diff. p
G

Figure 12: Discharge coefficient vs. Reynolds number
The Reynolds number can be calculated out of the identified velocities and the properties of the diesel fuel
and plotted vs. the discharge coefficient (Figure 12). Here it can be seen that the kink in the curves for injector B
is not located at fixed Reynolds number.
Conclusion
The validation of ten different spray models via a numerical investigation of a Mie-scattering measurement
campaign carried out that the models are only valid for none evaporative conditions. Although, some of the
models predicted the spray very good in some operating points none of the models turned out to be the best at all
operating points. The models of Wakuri, Naber and Siebers and Desantes did not meet the measured spray pene-
tration depth under the present testing conditions. The model of Arrgle had to be adjusted via a pre-factor which
had to be determined out of the measured data itself. The comparability of the models depends strongly on the
applied conditions and the injector itself. Other investigations showed that the exponent for time, chosen in the
most models with 0.5, is recommended. It also can be useful to shift the origin of this root of time curve in order
to get the first derivation valid at t=0 [15]. To describe the effects of evaporation, an additional term is necessary.
For the determination of the nozzle exit velocity the shifted root of time curve was fitted with a Matlab script
to the measured penetration depth vs. time data. To separate the propagation phase from the stationary spray
state an algorithm was used which is based on the standard deviation of the penetration. A strong increase in this
standard deviation is a marker for beginning evaporation.
The comparison of two injectors shows the influence of the complete injections system on the hydraulic be-
havior and the injection process. It was found that the discharge coefficient can significantly change in the opera-
tion range of an injector. By now, we are not able to say whether this phenomenon comes from the nozzle itself
9
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions
or the hydraulic layout of the whole injector. Inside the nozzle cavitation phenomena, such as a vena contracta
or a hydraulic flip, can affect the spray massively.
For future spray models it is useful to determine a discharge coefficient matrix over the entire operation
range and consider this coefficient in the spray model. This will especially be useful for injectors which show a
non-linear discharge coefficient.
10
24
th
ILASS Europe 2011 Comparison of current Spray models under high pressure and high temperature engine relevant conditions

Nomenclature
C
D
discharge coefficient [-]
C
V
velocity coefficient [-]
d
0
nozzle diameter [mm]
l
0
nozzle length [mm]
m mass [g]
p pressure [MPa]
Re Reynolds number [-]
S penetration depth [mm]
T temperature [K]
t time [s]
v velocity [ms
-1
]
v
measured
measured velocity [ms
-1
]
v
th
theoretical velocity [ms
-1
]

D
air entrainment coefficent [-]
temperature [C]
density [g/mm]
Subscripts
F fuel
G gas
References
[1] Y. Wakuri, Fujii, M., Amitani, T., Tsuneya, R., Bulletin J.S.M.E., 3 (1960) 123-130.
[2] G. Sitkei, Springer Verlag, Berlin, 1964.
[3] D.H.C. Taylor, B.E. Walsham, Proceedings of the Institution of Mechanical Engineers, Conference
Proceedings 1964-1970 (vols 178-184), 184 (1969) 67-76.
[4] J.C. Dent, SAE Paper, 710571 (1971).
[5] G. Lustgarten, ETH Zrich, Zrich, 1973.
[6] H. Hiroyasu, M. Arai, SAE Paper, 900475 (1990).
[7] J.D. Naber, D.L. Siebers, SAE Paper, 960034 (1996).
[8] J.M. Desantes, J. Arrgle, J. Pastor, SAE Paper, 980802 (1998).
[9] J. Arrgle, J. Pastor, S. Ruiz, SAE Paper, 1999-01-0200 (1999).
[10] S.S. Sazhin, G. Feng, M.R. Heikal, FUEL, 80 (2001) 2171-2180.
[11] B. Chehroudi, F.V. Bracco, SAE Paper, 880522 (1988).
[12] A.H. Lefebvre, Taylor and Francis, London, 1989.
[13] G.L. Borman, K.W. Ragland, McGraw-Hill, New York, USA, 1998.
[14] T. Vogel, M. Lutz, S. Iannuzzi, M. Wensing, A. Leipertz, Motorische Verbrennung, aktuelle Probleme und
moderne Lsungsanstze, Munich, 2011.
[15] T. Vogel, M. Lutz, M. Wensing, A. Leipertz, 23rd International Conference on Liquid Atomization and
Spray Systems, Brno, Czech Republic, 2010.


11

Você também pode gostar