Professional Documents
Culture Documents
Biodiesel Atomization
Biodiesel Atomization
Abstract
∗
Corresponding author
2
1. Introduction
NOTICE: this is the authors version of a work that was accepted for pub-
lication in the Journal: Fuel. Changes resulting from the publishing process,
such as peer review, editing, corrections, structural formatting, and other
quality control mechanisms may not be reflected in this document. Changes
may have been made to this work since it was submitted for publication. A
definite version was subsequently published in the Journal: Fuel, vol. 108,
pp. 758-770, 2013.
Spray formation is a process of crucial importance in improving internal combustion
engines, gas turbines and other combustion systems used in power plants and transport
applications. Small droplets are crucial in order to increase surface area and therefore
the mass transfer rates from the liquid to the gas phase. Generating a spray broadly in-
volves a series of atomization processes which for a simple cylindrical liquid jet injected
into air involves primary atomization (Dumouchel (2008); Faeth et al. (1995); Lefeb-
vre (1989); Wu et al. (1995); Tseng et al. (1992); Sallam et al. (2002)) and secondary
atomization (Guildenbecher et al. (2009); Faeth et al. (1995); Hinze (1955); Lefebvre
(1989)). Primary atomization is the break-up of a liquid jet into ligaments or fragments
(Faeth et al. (1995); Lefebvre (1989)), and secondary atomization involves the break-up
of droplets into smaller droplets (Guildenbecher et al. (2009)). The primary stage is
affected by a number of parameters which are fully described elsewhere (Faeth et al.
(1995); Dumouchel (2008)). In twin fluid atomization, which is of particular relevance
in gas turbine applications and applications which require a high degree of air entrain-
ment, the fluid flow properties of both the liquid jet and atomizing fluid become very
important (Lasheras and Hopfinger (2000)).
In a coaxial type air-blast atomizer, which is the focus of this contribution, primary
atomization is mainly dictated by the air and liquid physical properties, the blast air
velocity, the liquid jet velocity, along with certain geometric parameters which are ex-
tensively reviewed in the literature (Lasheras and Hopfinger (2000); Dumouchel (2008);
Lefebvre (1989)). Non-dimensional parameters of relevance include the liquid and air jet
3
Reynolds numbers (Rel and Rej ), the liquid jet Ohnesorge number (0hl ) and the exit
Weber number (W el ), provided in equations 1 to 4. Other non-dimensional parameters
which are important in describing the operation of an air-blast atomizer are detailed
elsewhere (Engelbert et al. (1995)).
ρl Ul Dl
Rel = (1)
µl
ρUj D
Rej = (2)
µ
µl
Ohl = √ (3)
ρl σl D
ρg (Ug − Ul )2 D
W el = (4)
σl
The last decades have emphasised the requirement for alternative energy sources
particularly in the transport sector, where combustible fuels will remain to be the prime
source of energy for the foreseable future. With the gradual advent of biofuels and
the increasing use of blends with a wide variation in physical properties, it is vital
to re-examine conventional atomization systems operating with these alternative fuels.
Combustible biofuels which originate from feedstocks such as soy, coconut, canola, palm
oils and other variants are gradually being considered (Allen and Watts (2000); Agarwal
and Chaudhury (2012); Allen et al. (1999); Llamas et al. (2012)).
Variations in fuel properties of biodiesels and organic fuels made from different feed-
stocks are expected to affect the atomization and combustion characteristics. Research
has been conducted to understand the physical properties of these fuels (Allen and Watts
(2000); Allen et al. (1999); Chhetri and Watts (2012)) as well as their spray and atom-
ization characteristics in a wide variety of spray experiments (Park et al. (2009, 2006,
2010); Kim et al. (2008); Al-Ahmad et al. (2009)) including constant volume chambers
(Agarwal and Chaudhury (2012)) and common-rail injectors (Kim et al. (2012)). More
fundamental studies have involved the study of secondary atomization of a monodisperse
4
droplet stream (Park et al. (2006)).
While work has clearly been conducted in improving our understanding of biodiesel
atomization, there is generally a lack of information on the atomization characteristics
of the constituent fatty acid methyl esters (FAMEs) of many of the biofuels investigated
in the past. Here, we investigate the spray characteristics of three FAMEs, namely C810
(saturated short chain length), C1214 (saturated medium chain length), and esterol 116
(methyl oleate, partially unsaturated long chain length), while we also compare with
ethanol. Ethanol, as with the biodiesels, is an oxygenated fuel, but has a lower viscosity
and surface tension than FAMEs.
Fatty acid esters made by alcohol transesterification have been recognized as diesel-
like fuels for compression ignition (CI) engines and kerosene-like fuels for gas turbine en-
gines (Llamas et al. (2012)). CI engines have operated successfully with pure biodiesels
and blends of biodiesels (Lapuerta et al. (2005)). Ethanol, considered a renewable
fuel, can be blended with gasoline (gasohol) in spark ignition (SI) engines (Pereira
et al. (2012)) and with diesel (diesehol) (Lei et al. (2012)) to operate CI engines and
homogeneous-charge-compression-ignition (HCCI) engines (Saxena et al. (2012)). Addi-
tionally, ethanol can also be a solvent for biodiesel production (Brunschwig et al. (2012)),
and therefore could be a sub-component in biodiesels which may have certain effects on
biodiesel combustion and spray characteristics.
Investigating FAMEs and ethanol fuels will allow for an isolated understanding of the
atomization and combustion characteristics of the consituent components of biodiesel.
The resulting knowledge will be useful for the development of improved computational
tools and for comparing with the performance of such fuels in real engines. In terms
of combustion, it is necessary to examine in simple yet representative experimental
setups, how issues such as the heat release and flame structures vary in a variety of
biodiesel driven flames, where similar analysis must be conducted as done previously for
conventional fuel spray flames (Karpetis and Gomez (2000); Oloughlin and Masri (2011,
2012); Prasad and Masri (2012)).
In this contribution, we concentrate on the atomization characteristics of non-reacting
5
sprays. A well characterised multiple stage atomizer in an air-blast mode is used in
order to qualitatively and quantitatively investigate the atomization characteristics of
these fuels. Measurements are made using phase Doppler/laser Doppler anemometry
in conjunction with a long distance microscope coupled to a high speed camera with a
significantly diffused high speed laser as the source of illumination. The imaging layout
has allowed for visualization of the break-up region. While primary and secondary at-
omization has been examined in the past, there is still a lack of quantitative information
regarding the primary atomization region or what is also referred to as the ‘dense spray’
region. While phase Doppler anemometry is commonly used in the primary atomiza-
tion region, it is quite inacurate due to the high data rejection rates present (Qiu and
Sommerfeld (1992); Tropea et al. (1996)). In this contribution, we extract quantitative
information from microscopic images that complements the PDA results such as block-
age area, and statistical quantities such as the probability of appearance of a particularly
shaped liquid fragment.
The paper is structured as follows. Firstly, the atomizer utilized for the experiments
is presented and described along with the experimental methodology and associated un-
certainties. Secondly, the test conditions are presented along with the physical properties
of the liquids used throughout the experiments. Droplet velocities and sizes conditioned
on the smallest droplets, which are taken to be representative of the gas phase while
also unconditioned results are shown as a function of radial position and downstream
location for all fuels. A brief description of the turbulent characteristics of the spray
are then provided for all fuels. A selection of representative break-up images is provided
followed by the description of the quantitative image processing procedure which is used
to give a measure of the density of the spray or liquid blockage area. Finally, the clas-
sification procedure which has been used to statistically describe the various atomized
liquid shapes is presented with results for all fuels.
6
2. Experimental Setup and Procedure
Figure 1 shows three dimensional views of the multiple stage atomizer utilized in
these experiments. The atomizer consists of an effervescent stage (zone 1) and an air
assisted stage containing airblast ports (port 2) and swirl ports (port 3). The efferves-
cent atomization mode offers a considerable reduction in droplet size at the expense of
a complex atomization process that involves interphase mixing upstream of the liquid
nozzle (part 1) (Sovani et al. (2001); Jedelsky et al. (2009)). This adds a level of compli-
cation which for the purpose of studying biodiesel atomization is not desirable. In these
experiments only the coaxially flowing airblast air of ports 2 is utilized as the air assisted
mechanism. The liquid nozzle diameter Dl fixed to part 1 is kept at a constant 0.5mm
for all experiments and the airblast nozzle diameter D of part 2 is fixed at 10mm. The
reader should note that this airblast geometry is somewhat different from a conventional
coaxial air assisted atomizer where the liquid injection nozzle is located on the same
horizontal plane as the air nozzle (Lasheras and Hopfinger (2000)). In this atomizer,
the liquid injection nozzle (part 1) is located upstream of the air nozzle for a number of
design reasons discussed elsewhere (Kourmatzis and Masri (2012)).
Figure 2 shows the diagnostic tools used to characterise the spray. A commercial laser
Doppler/phase Doppler anemometry system (TSI Model FSA 3500/4000) was used for
droplet velocity and size characterisation. The receiver was positioned in a 50 degree
forward scattering configuration. An Argon-ion laser feeds the two-channel fiber optics
assembly. This assembly transmits two pairs of beams with wavelengths 514.5 nm and
488 nm which are used for measuring the axial and radial components of velocity respec-
tively. A Bragg cell shifts one beam from each pair by 40 MHz to allow measurement
of velocity in the negative direction. Built-in probe volume correction (PVC) in the
software (FlowSizer) has been implemented to correct for lower detectability of small
droplets at the edge of the measurement volume. All parameter settings of the receiver
optic assembly were maintained constant for measurements throughout the length of
the jet in order to minimize any bias in the droplet size distribution relative to results
obtained at upstream locations. Typical validation rates past x/D=5 for all experiments
7
were over 90% in both channels. For cases A5 and A8 to be described in section 2.2,
this validation rate would drop to roughly 70-80% at the exit plane while for even higher
mass loadings such as cases A6 and A9 the rate could decrease even further, showing
the difficulty of probing the exit plane with a PDA system. The sampling rate varied
from 10-20 KHz at the exit plane centreline and decreased to typically 100-500Hz at the
edge of the jet for all cases and 10,000 samples were collected for each run.
A diode stack Nd-YAG laser operated at 532nm and 10KHz was used as the high
speed light source (Edgewave INNOSLAB model HD3011E). The average power used
in these experiments was 20W resulting in 2mJ/pulse. Two opal glass diffusing optical
components were used to scatter and randomize the light before crossing the measure-
ment volume and this was done in order to provide a uniform source of illumination.
A high speed camera (LaVISION, CMOS) was used in conjunction with a long dis-
tance microscope objective lens (QUESTAR, QM-100) in order to visualize a scale of
2.6mmx2.6mm in these experiments with a resolution of approximately 3.3µm. Finer
resolutions can be achieved however only through sacrificing the imaging window size.
Here, a large imaging window was vital in order to capture as many break-up events as
possible within the field of view.
8
mean diameter (SMD) has been tested by measuring the SMD of an ultrasonic nebulizer
with low velocity co-flowing air such that any further atomization is minimal. The
particular nebulizer has a manufacturer quoted SMD of 40µm with water. Measurements
of this ‘control’ spray have yielded SMDs that vary from 38 µm to approximately 45
µm depending on the co-flowing air velocity which is a deviation that ranges from 5-
10%. These errors however are only relevant in regions with high validation. Errors in
velocity will largely depend on what is known as ‘seeding bias’ fully described elsewhere
(McLaughlin and Tiederman (1973)).
Errors inherent with the flowmeters used in the experiment are of the order 5% and
1% for the liquid flow-rate and carrier air flow-rates respectively.
The boundary conditions of the airblast atomizer are shown in table 1. Five cases
were chosen so as to examine a range of different parameters including the effect of air
jet Reynolds number while also the effect of air/fuel ratio for all liquids. Table 1 is
valid for all three biodiesels and ethanol such that the volumetric flow-rate was varied
between each case in order to maintain a constant mass loading. Comparison of cases
4− >5− >6 reveals the effect of mass loading on the atomization behaviour of the
different fuels. Comparison of case 5− >8 or 6− >9 reveals the effect of air jet Reynolds
number for a fixed liquid loading. Table 2 shows the nomenclature used for the various
liquids used in the study such that case B1A5 for example, indicates the short chain
biodiesel operated under atomizing condition A5.
The results shall be structured as follows: Firstly, mean flow characteristics from
LDA shall be presented, followed by Sauter mean diameter results from PDA. A brief
presentation of the turbulent characteristics will then be displayed. The paper will then
proceed to present shadowgraph images, including the image processing methodology
and results.
9
3.1. PDA and LDA results
3.1.1. Mean flow characteristics
Radial profiles of the measured axial mean velocities are presented in Figs. 3 and 4
for various downstream locations in biodiesel and ethanol sprays. Results are shown for
the unconditional case where all droplet sizes are included as well as for a conditional
case where only small droplets (0-10µm) are considered. The conditioned results provide
an overall characterisation of the gas phase flow conditions of the atomizer for a selection
of fuels and cases, assuming that the smallest droplets closely follow the gas phase. The
Stokes numbers, St for these small 0 − 10µm droplets is at a maximum of approximately
St=0.05 at the exit plane where St is calculated as the ratio of the particle relaxation
timescale to integral timescale, fully described elsewhere (Gounder et al. (2012)). For all
LDA/PDA cases to be presented, the exit plane condition noted in all figures as x/D=0,
denotes the probe volume being located exactly 5mm above the atomizer surface. This
is due to physical limitations.
For all of the cases of figure 3 a clear decrease in the axial mean velocity is noted
at the exit plane when moving from outer radial locations to the centreline and this
is consistent with literature that has examined air-blast atomized sprays in the past
(Engelbert et al. (1995); Lasheras et al. (1998); Lasheras and Hopfinger (2000)). The
‘dip’ exists because the central liquid jet is still travelling at a lower velocity to that
of the surrounding coflowing airblast velocity, causing the full range of droplets in the
central core region to move slower.
This is particularly clear in high liquid flow-rate cases such as those of cases EA6 and
EA9 shown in figure 4. The reason for this is well understood (Lasheras and Hopfinger
(2000)) and occurs because in high mass loading cases, less of the liquid volume is
atomized at the exit plane, thus resulting in an average SMD which is much higher at
the centreline yielding a dense spray core which greatly affects the gas phase velocity. The
overall decay in gas phase axial mean velocity past x/D=0 is almost identical between
cases B1A5, B2A5 and EA5 as clearly seen in figures 3(a), 3(b) and 3(c).
While results from the various fuels conditioned on the smallest droplets (Figs. 3(a),
10
3(b) and 3(c)) are similar to each other, there is a difference when comparing the uncondi-
tional mean axial velocities of figures 3(d), 3(e) and 3(f ). Past x/D=0, the unconditional
results of Fig. 3 vary only slightly, and are within the bounds of experimental error.
However, at the exit plane (x/D=0) there is some variation in the axial mean velocity
amongst fuels. This variation shows that cases B1A5 (Fig. 3(d)) and EA5 (Fig. 3(f ))
have the lowest axial mean velocities compared to case B2A5 (Fig. 3(e)) which is ap-
proximately 20% higher than cases B1A5 and EA5 at the exit plane centreline. It is
further noted, that case B1A5 of Fig. 3(d) is showing a flatter velocity distribution at
the centreline when compared to the other two fuels. This is generally an indication
of better interphase mixing, which would generally be indicative of smaller droplets ap-
pearing in the spray. Images to be presented do indeed suggest smaller droplets in cases
B1A5 and EA5 when compared to B2A5, however this ‘flatness’ is not observed in figure
3(f ) where it would also be expected. The precise reason for this is unknown however
results in this location will be severely affected by the presence of ligaments which are
not included in the PDA measurement due to the limitations of the system. For a further
analysis of the exit plane region, images shall be analysed and discussed in subsequent
sections.
Figure 4 shows the velocity distributions for cases conditioned on the smallest droplets
(d=0-10) as well as for unconditioned results. These plots relay the effect of Reynolds
number and mass loading as opposed to the effect of fuel properties, which was the
purpose of figure 3. Figures 4(a) and 4(d) may be compared to figures 3(b) and 3(e)
respectively. One can clearly see how at the lower mass loadings of case B2A4, the axial
mean velocity at the exit plane becomes flatter when compared to B2A5 due to the
greater degree of interphase mixing. However, downstream results are largely unaffected
which is also consistent with dilute spray results (Chen et al. (2006); Gounder et al.
(2012)). The same conclusions can be drawn by comparing figures 3(c) and 3(f ) to
figures 4(b) and 4(e) respectively, where it is seen that the higher loading case of EA6
results in a much more distinct drop in axial mean velocity at the centreline.
Figures 4(b) and 4(e) may be compared to figures 4(c) and 4(f ) in order to examine
11
the effect of the Reynolds number for a fixed high mass loading. While the overall mean
velocity increases when going from case A6 to A9, the difference in maximum velocity
(@ ∼ r/D = 0.4) and minimum velocity (@ ∼ r/D = 0) at the exit plane centreline has
remained roughly constant. This indicates that the overall mean convective properties
have changed but the degree of interphase mixing has remained relatively unchanged,
given the similar momentum characteristics of the dense core. Again, this must be
studied further through examination of images and not solely from the LDA data.
12
imal in the downstream region. This is an important result as it indicates that any
differences in spray combustion characteristics between two fuels as different as B1 and
B3 will not be attributed to the spray dynamics.
The same general conlusions may be reached by examining figures 6(d) and 6(e)
which show similar droplet size characteristics past x/D=5. The results here are of
direct relevance to any air assisted atomization system, and will form the benchmark
to reacting spray experimentation to be conducted by the authors in future work. It is
very important to note that while droplet sizes are similar amongst the various fuels in
this air assisted system, this is only because all of the atomization occurs upstream of
x/D=5, mainly due to the significant level of air assistance. Therefore, in atomization
systems with rapid break-up times such as this one, the fuel physical properties will not
be as important as chemical properties when studying combustion characteristics. The
reader is directed to the literature (Faeth et al. (1995); Dumouchel (2008)) for work on
single fluid pressure atomizers which fully describe the effect of fuel physical properties
on break-up times and performance.
The apparent similarity amongst the different fuels can also be explained using a
simple scaling analysis. Assuming no air assistance, then the ligament break-up time
may be estimated through equation 5 (Faeth et al. (1995)). The representative diameter
scale (SMD) in the equation is taken to be the liquid orifice diameter equal to 0.5mm.
The Sauter mean diameter SMD at the exit plane based on the PDA results is unreliable,
which can be confirmed through examination of the shadowgraph images. Calculating
τlig for the various fuels yields times of the order 1.6-2 ms for the various fuels. This
can be compared to the time taken for a liquid packet to be advected from x/D=0
to x/D=5, estimated here as τmean = Uav /5D where Uav is the average of the mean
centreline unconditional velocity at x/D=0 and x/D=5. Calculating τmean for case EA5
for example, yields 1.1ms while for case B2A5 τmean is equal to 0.9ms. Therefore, based
on this theoretical estimate τmean ∼ τlig between x/D=0 and x/D=5. This however does
not take into account air assistance, which will greatly accelerate the break-up process
causing τlig < τmean . Therefore, it is clear that this theoretical estimate suggests that
13
the biodiesels and ethanol must all fully break-up within 5 diameters. The experimental
data not only confirms this, but also confirms that the droplet sizes they produce are
extremely similar.
1/2
ρl SM D3
τlig = (5)
σl
The overarching conclusion reached from the PDA results is that in the downstream
region, the droplet sizes are extremely similar, as are the momentum decay and tur-
bulence intensity characteristics. Exit plane results also showed little variation. As a
complement to the PDA results, a high speed imaging study has been undertaken in
an attempt to gather further information on what occurs in the primary atomization
region, where most of the differences in atomization behaviour would be expected for
liquids of different surface tension.
14
3.2.1. Qualitative discussion
Figure 8 shows typical instantaneous shadowgraph images for the spray of case A5
at the exit plane, revealing clearly how much information a PDA measurement would
exclude in that area. Each row is for a different liquid labelled on the figure, while each
column is for a different point in time separated by 100 µs. Through the images one can
clearly distinguish between individual droplets and the unbroken portions of liquid such
as ligaments and highly deformed droplets. All liquids resulted in qualitatively the same
atomization structure. However, the more viscous liquids such as B2 and B3 seemed to
result in a larger population of longer, stretched ligaments. This is due to the higher
surface tension of B2 and B3, and shows how even at the high strain rates which prevail
here, the viscous liquids maintain a large portion of unbroken liquid at the exit plane of
the air nozzle.
Concentrating on the row of images for B3 of figure 8, one can track the evolution
of a biodiesel ligament travelling from the bottom to the top of the field of view. The
ligament remains intact over the distance probed, showing that the breakup time of the
liquid packet is greater than 100 µs which is the interframe separation.
Ligaments of length such as that of the one displayed for case B3 were uncommon
with biodiesels B1 and B2 and ethanol. Figure 9 shows representative images for the
other two cases imaged for liquid B3, particularly cases A4 and A8. Case A4 generally
resulted in a small population of undeformed droplets and was a well atomized spray,
thus resulting in an adequate validation rate as described in section 2. Case A8 of figure
9 as with B3A5 of figure 8 resulted in a significant population of stretched ligaments
and unbroken liquid and this shall be revisited quantitatively in the sections to follow.
Comparing typical instantaneous images of liquids B1 and E for case A5 did not reveal
any striking differences, however this shall also be investigated further upon statistical
analysis of the images.
15
dimensional structure of the spray. The objective is to be able to identify and classify
the fluid elements which are rejected by the PDA system, as well as track their evolution
leading to the final formation of droplets. Some initial attempts at processing such
images are made here, where two different image processing techniques are applied; one
automated, and one manual, each one providing different information complementary
to the PDA data. The reader must take note that these measurements represent line-
of-sight images of a three dimensional process and therefore will inherently contain a
number of uncertainties, including defocused objects, objects hidden behind liquid on
the imaging plane, while also information completely outside the spatial field of view both
in terms of width and depth. Future work should be attempted to address each of these
innacuracies however this study presents an initial investigation into alternative ways of
investigating atomizing sprays. The methods are fully described in the subsections to
follow.
Automated technique
The automated image processing method determines the amount of liquid that blocks
the field of view at the centreline of the atomizer. This provides one direct quantitative
measure of the degree of atomization at a given location. For a given initial jet Reynolds
number and liquid jet mass flow-rate and thus roughly fixed initial level of dispersion,
any change in the amount of liquid present in the field of view is a result of atomization.
More atomization will result in smaller droplets forming, which will have lower inertia
and thus a larger propensity to disperse and leave the field of view. Given however
that the backlit illumination technique is a line-of-sight method of visualization, liquid
packets that are defocused are also included in the image. For that reason, thresholding
has been applied to the image prior to processing. Figure 10 shows an example of
the thresholding procedure. The left frame of figure 10 shows the raw image displayed
by the software DaVIS, showing darkened regions in the background which represents
defocused liquid. Applying a threshold of 200 counts for example, removes some of that
background, however not fully as shown in the middle frame. A threshold of 100 counts
removes all of the background, while it also filters out some liquid which is present at the
16
imaging plane. Some structures which are easily distinguishable such as the ligament
circled in the images will be filtered out of the final processed result, given that portions
of the ligament are not located on the imaging plane. While this results in some loss of
information it is consistent amongst all images chosen in the processing methodology,
and it also removes spurious defocused information entirely.
Figure 11 shows an example of the ratio of liquid area to total area for cases EA5 and
EA6 for 1000 images. The calculation result shows the intermittency in blocked area
and how the higher loading case of EA6 clearly results in events where liquid has blocked
up to 25% of the field of view (note the different scale for Figs 11(a) and 11(b)). Such
a high blockage is not recorded for lower loading cases such as EA5. This calculation
procedure is repeated at the centreline for 1000 images from x/D=0 to x/D=5 for all
liquids and a variety of cases. The mean value provides a statistical quantity related to
the degree of blockage.
Figure 12 shows the mean Arealiquid /Areatotal ratio for all liquids for cases A5 (12(a))
and A8 (12(b)) showing the effect of Reynolds number. Figure 12(a) clearly shows how
with downstream location there is an expected decrease in total area blocked. However
Fig. 12(a) statistically reveals a much larger blockage at the exit plane for biodiesel B3
while the other 3 liquids are generally quite similar. The rate of change of the blocked
area with downstream location is also different amongst fuels, however by approximately
x/D=4 it is clear that all liquids have converged onto a similar blockage area ratio. Of
interest to note is the much steeper decrease in blockage area ratio with x/D for case
B3 of figure 12(a), between x/D=0 and x/D=1 when compared to the other cases. This
indicates the presence of liquid which breaks up more rapidly for case B3, and may be
related to the presence of ligaments oriented perpendicular to the flow, which break up
into droplets and then disperse. The presence of long ligaments in case B3 shall be
revisited in the following subsection. The fact that the blockage area seems to converge
to a similar value past x/D=4 is in direct agreement with the PDA data which suggests
that past x/D=5 the overall droplet size and spray dynamics are effectively the same.
Clearly, the overall statistical break-up time between liquids B3, B2 and B1 is dif-
17
ferent given the different slopes of the area ratio history shown in Fig. 12, indicating
a strong effect of liquid physical properties on the degree of atomization upstream of
x/D=5. This, however, is not the case when comparing liquid B1 to ethanol which has
an almost identical evolution of blockage area with downstream position. Given the
similarity of the physical properties of ethanol and B1 as may be seen through table 2
this is a comforting result, and acts as some confirmation that the processing method
is able to quantitatively capture differences in the atomization characteristics due to
liquid physical properties. Figure 12(b) shows the mean area ratios for case A8 for all
fuels. The overall trend is similar with all cases again converging onto a similar blockage
ratio at roughly x/D=5, while showing some significant differences in the change in the
mean value. Cases B1 and ethanol again are almost identical, however fuel B3 seems
to have an overall larger blockage area when compared to the same liquid for case A5.
This is counterintuitive as a higher Reynolds number should result in a higher degree of
break-up. Fuel B2 results in a higher blockage area at the exit plane for case A8 when
compared to case A5 although past x/D=1 the blockage area history is similar to case
A5 for fuels B2,B1 and E. The result suggests that even for the higher Reynolds number
of case A8, upstream of x/D=5 there is still a significant difference in the atomization
behaviour of biodiesel B2 and B3 when compared to the other two fuels.
This approach provides at least some quantitative measure of atomization at loca-
tions where PDA rejects certain information, and may provide a good reference for the
validation of computations in the primary break-up region. Here, we have suggested the
simple calculation procedure as a way of comparing biodiesel atomization behaviour,
however further rigorous testing of the processing methodology should be carried out in
the future so that it can be applied in a variety of spray scenarios.
A limitation of this approach is that it does not provide statistical information on
the break-up structure and it cannot include defocused objects. In the next section, we
utilize a manual processing method in order to compare in further detail, the atomization
behaviour of the different fuels.
Manual classification technique
18
After manual examination of multiple shadowgraph images of the various fuels, four
generic liquid structures have been identified for all fuel cases. These are defined as
follows: Structure A is a long ligament, or a stretched portion of liquid significantly
larger than the liquid nozzle diameter such that its length is at least 20% of the field of
view. Structure B is a short ligament, or a stretched portion of liquid which is no longer
than 20% of the field of view, this corresponds to a size of 1.04Dl , or 520µm. Structure
C is a deformed droplet. Structure D is a large unbroken liquid volume, defined as a
volume of liquid with a characteristic length at least the size of the liquid orifice diameter
Dl , which has an aspect ratio greater than approximately 0.3. Structure D is therefore
clearly distinguished from A, based on the aspect ratio. Figure 13 shows an example
image with the structures defined being clearly identified. The total number of objects
falling into each category A,B,C or D have been counted and normalized by the total
number of all objects counted. The resulting quantity is the probability of a particular
liquid structure appearing in the spray. This classification technique was initially applied
to twenty, fifty, seventy and one hundred images for case A5, for all liquids. This was
conducted in order to observe if any significant change in probability trend occured with
the number of images sampled.
Figure 14 shows the probability of an event vs. the jet exit Weber number of equation
4 as a function of the number of images. The jet exit Weber number is a key indicator
of primary atomization in airblast atomizers, and it incorporates the density, viscosity,
and surface tension of the various liquids acting as a good non-dimensional indicator
for this particular study. Clearly, the overall probabilities and trends in probability of
a given event do not vary greatly as a function of the number of images. However it is
clear particularly through figures 14(b) and 14(c) that 20 images are insufficient for a
cacluation as they deviate the most from the statistics calculated based on more samples.
Figure 15 shows the probabilities vs. exit Weber number (W el ) calculated based on
100 images for cases A5 and A8. The probability of a long ligament appearing varies
from 20% to under 5% for case A5 as seen in in Fig. 15(a) while this decreases to a
range of 9% to approximately 2% for case A8 of Fig. 15(a), clearly showing how the
19
larger Reynolds number changes both the probability of long ligaments as well as the
rate of decrease in probability of long ligaments with Weber number. This indicates that
at the higher Reynolds number case of A8 the probability of long ligaments appearing is
less variable with liquid physical properties, due to the higher impact of air jet velocity
on atomization. As with figure 15(a), figure 15(b) shows a decrease in probability of
events with exit Weber number, clearly indicating an effect of physical properties. The
probability of a short ligament appearing for case A8 is higher than case A5, ranging
from 45% to 32% compared with 35% to 25% for case A5. This suggests that the highest
jet Reynolds number case results in a higher frequency of short ligaments. In contrast
to figures 15(a) and 15(b), figure 15(c) shows an increase in the probability of an event
with Weber number. Clearly, deformed droplets are most likely to appear in these sprays
with probabilities ranging from 40% to 65% for cases A5 and A8. The curves of figure
15(c) further suggest that the probability of deformed droplets appearing are not greatly
influenced by the jet Reynolds number. Finally, the probability of a large unbroken liquid
volume occuring is relatively unchanged with exit Weber number for a given Reynolds
number and as with figure 15(c) the curve is a shift to the right with little change in
slope or magnitude. Therefore the appearance of long and short ligaments seem to be
the most influenced by Reynolds number.
This method of classifying the level of atomization, while labor intensive, is generally
consistent with the liquid physical properties while also with the flow characteristics. We
suggest such a statistical classification of a spray as a complementary indicator of spray
atomization in regions where PDA rejects non-spherical droplets. In the future, the
technique should be automated in order to allow for more images to be included in the
analysis, and for improved accuracy.
4. Conclusions
Two experimental techniques, PDA and high speed microscopic imaging have been
utilized to investigate the atomization behaviour of three biodiesels (fatty acid methyl
esters of short, medium and long chain length) and ethanol. A simple automated pro-
20
cessing technique has been used as a measure of spray blockage at a particular location
while a manual processing method has been employed with four key structures being
statistically identified. The probability of ligaments, deformed droplets and unbroken
liquid volumes has been calculated yielding consistent information while also details on
the atomization behaviour of these biodiesels when compared to ethanol. The work
presented here provides a systematic study into the atomization characteristics of three
fairly well defined biodiesels with respect to ethanol, a commonly used biofuel, and has
a provided a detailed account of the microscopic spray structure as well as an analysis of
global quantities. The study has yielded a number of conclusions pertinent to the study
of the atomization behaviour of these fuels:
• PDA results reveal that the structure of this spray is similar to a conventional
airblast atomization geometry.
• Momentum decay and droplet size characteristics are very similar amongst the
three biodiesels and ethanol past x/D=5.
• Liquids B1 and ethanol yield very similar blockage area results suggesting that their
atomization characteristics are similar while also confirming the consistency of the
image processing technique given the very similar surface tension and viscosity of
fuels B1 and E.
21
and short ligaments appearing at the exit plane, an increase in the probability of
deformed droplets appearing, as well as a generally unchanged probability of large
unbroken volumes appearing.
• The higher surface tension (low Weber number) biodiesels, particularly B2 and
B3 result in a significant amount of the breakup being in the form of stretched
ligaments.
5. Acknowledgments
The work is funded by the Australian Research Council. The authors thank Mr.
Priyadarshi Pandey and Dr. Mrinal Juddoo for assistance with the experiments, Mr.
Kumaresan Sivapalan for assistance with the solid modelling of the atomization device
and Dr. Vinayaka Nakul Prasad for insightful comments throughout the analysis of the
work.
22
References
L. Tseng, G. Ruff, G. Faeth, Effects of gas density on the structure of liquid jets in still
gases, AIAA Journal 30 (1992) 1537–1544.
K. Sallam, Z. Dai, G. Faeth, Liquid breakup at the surface of turbulent round liquid
jets in still gases, International Journal of Multiphase Flow 28 (2002) 427–449.
J. Lasheras, E. Hopfinger, Liquid jet instability and atomization in a coaxial gas stream,
Annual Review of Fluid Mechanics 32 (2000) 275–308.
23
A. Agarwal, V. Chaudhury, Spray characteristics of biodiesel/blends in a high pressure
constant volume spray chamber, Experimental Thermal and Fluid Science 42 (2012)
212–218.
C. Allen, K. Watts, R. Ackman, Predicting the surface tension of biodiesel fuels from
their fatty acid composition, Journal of the American Oil Chemists’ Society 76 (1999)
317–323.
S. Park, H. Kim, H. Suh, C. Lee, A study on the fuel injection and atomization charac-
teristics of soybean methyl ester (sme), International Journal of Heat and Fluid Flow
30 (2009) 108–116.
S. Park, H. Suh, C. Lee, Nozzle flow and atomization characteristics of ethanol blended
biodiesel fuel, Renewable energy 35 (2010) 144–150.
24
H. Kim, S. Park, H. Suh, C. Lee, Atomization and evaporation characteristics of biodiesel
and dimethyl ether compared to diesel fuel in a high-pressure injection system, Energy
and Fuels 23 (2012) 1734–1742.
P. Pereira, M. Marra, R. Monz, E. Richter, Fast batch injection analysis system for on-
site determination of ethanol in gasohol and fuel ethanol, Talanta 90 (2012) 448–457.
J. Lei, L. Shen, Y. Bi, H. Chen, A novel emulsifier for ethanol-diesel blends and its effect
on performance and emissions of diesel engine, Fuel 93 (2012) 305–311.
S. Saxena, S. Schneider, S. Aceves, R. Dibble, Wet ethanol in hcci engines with exhaust
heat recovery to improve the energy balance of ethanol fuels, Applied Energy 98
(2012) 448–457.
25
H.-H. Qiu, M. Sommerfeld, A reliable method for determining the measurement volume
size and particle mass fluxes using phase-doppler anemometry, Experiments in Fluids
13 (1992) 393–404.
J. Gounder, A. Kourmatzis, A. Masri, Turbulent piloted dilute spray flames: Flow fields
and droplet dynamics, Combustion and Flame 159 (2012) 3372–3397.
26
A4 A5 A6 A8 A9
Ujet (m/s) 64 64 64 74 74
mair (kg/min) 0.36 0.36 0.36 0.42 0.42
mliquid (kg/min) .02 .04 .07 .04 .07
Rejet 42666 42666 42666 49333 49333
B1 B2 B3 E
Fuel Short-chain Medium-chain Long-chain Ethanol
(C810) (C1218) (C1618) C2H60
VCC-91208 VCC-91216 Esterol 116
Density ρ (kg/m3 ) 877 871 873 789
Viscosity µ (Pas) .00199 .00434 .00489 .0013
Surface tension σ (N/m) .025 .033 .044 .024
Ohl .019 .036 .035 .014
Table 2: Liquids used for the investigations including physical properties relevant to atomization, and
fuel codes given by supplier
27
Figure 1: Schematic of multiple stage atomizer housing with isometric cutaway view, where part 1 shows
the liquid nozzle, part 2 shows the air-blast nozzle, zone 1 shows the effervescent stage (not used), port 2
shows one of the air-blast inlet ports, and port 3 shows one of the swirl inlet ports.
28
Figure 2: Schematic of experimental measurement techniques including laser Doppler/phase Doppler
anemometer and microscopic high speed imaging system where E is the emmitter and D is the detector
of the LDA/PDA system (The two techniques were not conducted simultaneously)
29
(a) B1A5 d=0-10 (b) B2A5 d=0-10 (c) EA5 d=0-10
Figure 3: Unconditional mean axial velocities (d-f) and mean axial velocities conditioned on droplet
sizes from d=0-10 µm (a-c) plotted vs. r/D for biodiesels B1 and B2 and ethanol for case A5.
Figure 4: Unconditional mean axial velocities (d-f) and mean axial velocities conditioned on droplet
sizes from d=0-10 µm (a-c) plotted vs. r/D for biodiesel B2 and ethanol showing the effect of loading
for fixed Reynolds number (B2A4− >B2A5) and the effect of Reynolds number for fixed mass loading
(EA6− >EA9).
30
(a) EA5 (b) EA6
(c) EA9
Figure 5: Sauter mean diameter vs. r/D for ethanol, showing the effect of Reynolds number for fixed
mass loading (EA5− >EA6) and the effect of mass loading for fixed Reynolds number (EA6− >EA9).
31
(a) B1A4 (b) B2A4 (c) B3A4
Figure 6: Sauter mean diameter vs. r/D for all three biodiesels and ethanol for a selection of cases.
Figure 7: Turbulence intensity conditioned on diameters d=0-10 µm (a-c) plotted vs. r/D for a selection
of cases.
32
Figure 8: Selection of images in time (left to right) showing typical atomization events at the exit plane
centreline for the three biodiesels (B1,B2 and B3) and Ethanol (E) for case A5. Images in each row are
time sequences separated by 100µs
Figure 9: Selection of images in time (left to right) at the exit plane centreline showing typical atomization
events for Esterol 116 (B3) for cases A4 and A8
33
Figure 10: Sample image (left) utilized to illustrate thresholding procedure at a threshold of counts=200
(middle) and counts=100 (right)
34
(a) EA5
(b) EA6
Figure 11: Sample calculation of blockage area history vs. number of images at the exit plane centreline
(each image separated by 100µs) showing the effect of loading for ethanol.
35
(a) Case A5
(b) Case A8
Figure 12: Mean blockage area vs. x/D for all fuels for cases A5 and A8.
36
Figure 13: A selection of images showing the manual break-up classification procedure, where A is a long
ligament (length> 1.02dliquid ), B is a short ligament (length< 1.02dliquid ), C is a deformed droplet and
D is an unbroken liquid volume
37
(a) Long ligament appearance (A) (b) Short ligament appearance (B)
Figure 14: Probability of a particular liquid volume appearing in an image vs. exit Weber number (W el )
as a function of the number of images used for the calculation for case A5.
38
(a) Long ligament appearance (A)
Figure 15: Probability of a particular liquid volume appearing in an image vs. exit Weber number (W el )
for case A5 and A8 using 100 images.
39