Atomization and Sprays, 26 (10): 1009–1030 (2016)

GENERATING PLANAR STATISTICS FROM

POINT MEASUREMENTS IN A SPRAY
Kyle M. Bade∗ & Rudolf J. Schick

Spray Analysis and Research Services, Spraying Systems Co., Wheaton, IL

60187, USA

∗
Address all correspondence to: Kyle M. Bade, E-mail: Kyle.Bade@Spray.com

Original Manuscript Submitted: 4/21/2015; Final Draft Received: 9/21/2015

Methods used to generate statistical values of planar spray characteristics, such as mean drop size and
velocity, from discrete measurement points are established and evaluated. Round, flat, and multiori-
fice air-atomized sprays are investigated experimentally, representing common but distinctly different
sprays. The details of each spray plume are overresolved with an extensive matrix of measurement
points, which are then down-sampled to evaluate the minimum number of required points and loca-
tions to calculate accurate planar spray statistics. Weighted-average values are generated using the
local flux and discrete area at each measurement point, which provide holistic spray characteristics
using relevant physical characteristics of the spray plume. It is found that measurements should be
performed at approximately each 10% interval of the overall spray plume size to achieve errors less
than 5%. Also, the measurements should be performed along one axis for an axisymmetric round
spray, along both the x- and y-axes for a flat spray, and along a single axis from the center to edge for
a multiorifice air-atomized spray for drop size, but along both the x- and y-axes for velocity. Finally,
the methods are applied toward the combination of drop size distributions acquired at many points to
generate a representative combined drop size distribution for a spray cross section.

KEY WORDS: spray characterization, planar statistics, processing methods

1. INTRODUCTION

Significant effort has been made by many researchers to establish meaningful statistics to
represent the drop size and velocity characteristics of chaotic and complex spray plumes.
For example, in the well-recognized text by Lefebvre (1989), and originally proposed by
Mugele and Evans (1951), definitions are presented for the arithmetic mean, Sauter mean
diameter, and many other statistical values that can be generated from well resolved
drop size distributions. These distributions may be collected at a point within a spray
or throughout a spray plume. While point measurement methods are established and ac-
curate, methods to compare sprays holistically are not well developed. Many methods
and instruments, some of which are reviewed in texts by Albrecht et al. (2003); Dodge
et al. (1987); Tishkoff et al. (1984), may collect these distributions; however, a concise

1044–5110/16/$35.00 ⃝
c 2016 by Begell House, Inc. 1009
1010 Bade & Schick

demonstration of the combination of multiple point measurements into single character-

istic value for a spray plume has not been developed. Planar spray characteristics may
be directly collected by techniques such as interferometric particle imaging (IPI) (see
Albrecht et al. (2003)), but these methods often suffer from limited ability to measure
within a moderate to very dense spray plume. Furthermore, a significant advantage of
point measurement systems, beyond the ability to characterize dense sprays, is the abil-
ity to collect flux information. Through this information, regions of high flux can be
treated as more influential to resulting planar statistics. This is not possible with imaging
techniques that only provide instantaneous or ensemble (spatial) spray measurements.
The current investigation aims to use the flux information provided by point measure-
ments from a phase Doppler interferometer (PDI) to generate flux sensitive planar spray
statistics. Furthermore, an investigation is carried out to examine the minimum number
of measurement points required to calculate accurate planar mean values for multiple
candidate sprays that cover a range of typical nozzle designs and builds on the previous
work by Bade and Schick (2013).
A comprehensive review of characterization techniques developed for SAE Inter-
national (2007) and subsequent papers by Hung et al. (2008) and Gandhi et al. (2008)
provide a list of the available measurement techniques for spray characterization, and
provide an impressive list of recommended tests to characterize a fuel injection spray,
but do not evaluate the accuracy of these methods or present methods to use point mea-
surements for holistic comparisons. More recently, Falser and Parrish (2015) provided a
comprehensive review on many spray measurement instruments, the benefits and draw-
backs, and describe where in the the spray they are best used. These reviews of how to
measure sprays are quite instructive but do not specify where to measure within a spray
pattern, the number of measurements that are required (for point measurements), or how
to collect the resulting information into holistic values for comparison.
Published demonstrations of postprocessing of point measurement techniques to ob-
tain planar statistics are rare. Bade and Schick (2008) provide an investigation of PDI
measurements and LSI measurements but do not combine the results into planar values.
Notably, Chapple and Hall (1993) investigate a hydraulic flat spray nozzle for the agri-
cultural industry and collect measurements at 225 points within a planar spray plume
using a phase Doppler particle analyzer (PDPA) instrument to determine the optimal
measurement locations. This is similar to the current efforts, except that only straight
averages of the collected measurement points are assessed, whereas weighted averaging
methods are also developed and implemented in the current study. Tuck et al. (1997) also
investigate a hydraulic flat spray and perform measurement scans throughout an entire
quadrant of the spray area and use a straight average of the entire scan, a process they
acknowledge as very time consuming.
The results of the present study provide clear recommendations, supported by the
data, on where to measure with point measurement instruments and the number of mea-
surement points that should be acquired. Furthermore, methods are developed to com-

Atomization and Sprays

Planar Statistics from Point Measurements 1011

bine these reduced sets of measurements into reliable holistic quantities. Presumably,
these methods could be applied to experimental data, but also to collapse the highly
detailed data fields acquired through computational efforts for generalized comparisons.

2. EXPERIMENTAL METHODS

2.1 Nozzles

Measurements were acquired for three nozzle types: (1) a single-orifice, hydraulic, full-
cone, round spray; (2) a single-orifice, hydraulic, flat spray; and (3) a multiorifice, air-
atomized spray; all nozzles are standard Spraying Systems Co. products and are refer-
enced herein as the round, flat, and multi orifice nozzle spray patterns, respectively.
The investigated round spray pattern was generated by a FullJet 1/8G-1 pressure
swirl nozzle and produced a round, axisymmetric, full-cone spray pattern. This nozzle
was operated at a liquid pressure of 2.75 bar, with a flow rate of 710 mL/min. Figure 1
provides an ensemble average laser sheet imaging (LSI) testing result at the z = 100 mm
plane with this nozzle and includes an overlay of the 43 discrete measurement points
collected along a single spray axis with the PDI instrument (see Section 2.2 for a com-
plete description of the LSI instrument and methods). Note that the imperfections in the
round spray pattern are generated by the swirl element within the nozzle; the spray does
become more axisymmetric at greater spray distances.
The flat spray was generated by a UniJet TPU650050-TC nozzle and provided an
attractive case given that this nozzle was previously investigated in detail by Bade and
Schick (2011). The nozzle was operated at a liquid pressure of 4.0 bar, resulting in a
flow rate of 225 mL/min. This flat spray nozzle provides a volume distribution that is
fairly symmetric about the x- and y-axes, but not axisymmetric. Figure 2 demonstrates

80
I /I m a x
60 1
40 0.5
y (mm)

20 0
0
−20
−40
−60
−80
−80 −60 −40 −20 0 20 40 60 80
x (mm)

FIG. 1: LSI result for the round spray FullJet 1/8G-1 nozzle with overlaid PDI test
points.

Volume 26, Issue 10, 2016

1012 Bade & Schick

30
I /I m a x
1
20
0.5
10

y (mm)
0
0

−10

−20

−30
−50 −40 −30 −20 −10 0 10 20 30 40 50
x (mm)

FIG. 2: LSI result for the flat spray UniJet TPU650050-TC nozzle with overlaid PDI
test points.

this symmetry at the z = 50 mm plane, and also includes an overlay of the 273 discrete
measurement points acquired over the entire spray plume during PDI testing.
The multiorifice, internally air-atomized nozzle provided an additional, more com-
plex spray plume shape for further analysis of the optimal spray measurement point
location subsets for interrogation when generating planar spray statistics. Figure 3 pro-
vides the LSI result of the average spray plume at the z = 178 mm plane at 2.4 bar water
pressure and 2.67 bar air pressure, which resulted in a water flow rate of 136 mL/min

150
I /I m a x
1

100 0.5

0
50
y (mm)

−50

−100

−150
−150 −100 −50 0 50 100 150
x (mm)

FIG. 3: LSI result for the multiorifice air-atomized SU26B nozzle spray, with overlaid
PDI test points.

Atomization and Sprays

Planar Statistics from Point Measurements 1013

and an air flow rate of 0.0756 standard mL/min. For this study, the complex shape of the
spray plume generated by a single orifice was investigated. Figure 3 includes an overlay
of the array of 453 discrete measurement points overtop the interrogated plume.

2.2 Laser Sheet Imaging

LSI measurements were conducted for each nozzle to provide high-spatial-resolution,

xy planar contours representative of the spray distributions generated by each nozzle.
A LaVision SprayMaster system was used to collect 200 instantaneous images for each
spray. The general setup of the LSI instrument including the laser, camera, and nozzle
orientation is demonstrated in Fig. 4. The camera was located approximately 40◦ upward
from the laser sheet and 2 m from the spray center, which allowed each spray pattern
to appear relatively in focus. The images were also corrected for the off-axis camera
perspective distortion using images of a regularly spaced grid.
The ensemble-averaged planar contour results from the LSI testing at z = −100,
−50, and −178 mm from the nozzle exit orifices are provided for the round, flat, and
multiorifice sprays in Figs. 1, 2, and 3, respectively, with contour lines drawn at 5%
intervals of the maximum level. The relative light intensity, I, in these results is scattered
according to Mie theory and is representative of the spatial distribution of the surface
area of all droplets passing through the laser sheet, that is, the average surface area
distribution. While these distributions are not strictly representative of the spray volume
distribution, these data were used to guide the selection of the locations and number
of PDI test points in each spray with an attempt to acquire many measurement points
within each 5% contour of each distribution. As demonstrated in Figs. 1–3, there are
measurement points within every 5% contour line throughout each of the xy planar
spray plumes. The spatial resolution of the xy plane LSI images was 0.56, 0.24, and
0.56 mm/pixel for the round, flat, and multiorifice sprays, respectively.

z
y x

FIG. 4: LSI setup demonstrating the (left) laser, (right) camera, and nozzle orientations.

Volume 26, Issue 10, 2016

1014 Bade & Schick

2.3 Phase Doppler Interferometry

PDI measurements were acquired, using an Artium Technologies PDI-MD-200 instru-

ment, throughout a single xy plane of each spray plume; see Fig. 5 for the general setup
of the PDI instrument. Measurements were acquired at the points shown in Figs. 1–3,
to provide measurements of droplet size and velocity as well as local volume flux. As
a matter of practicality, all PDI measurement points were acquired at constant intervals
from the nozzle center point; 10,000 samples were acquired at each point to provide
converged values (±1%). For the round spray, measurements were acquired at 3 mm
increments along the x-axis only due to the nominal axisymmetry of this spray. For the
flat spray, measurements were acquired at 3 mm increments in both x and y over the
entire spray pattern. Finally, for the multiorifice spray, measurements were acquired at
every 3 mm in x and every 6 mm in y throughout one of the six plumes. The accuracy
of the PDI measured droplet size and velocity was rigorously investigated by Bade and
Schick (2011), and shown to be fairly insensitive to setup parameters when using the
Artium AIMS software auto-setup feature. Similar acquisition methods were used as
those outlined as optimal by Bade and Schick (2011), with the transmitter/receiver lens
combination of 1000/500 mm and the receiver located at a 40◦ off-axis position, which
allowed for a sufficient drop size measurement range and resolution. Bade and Schick
(2011) also provide an extensive investigation of the accuracy of volume flux measure-
ments using the PDI instrument across the major axis of a flat spray plume; the results
are found to be accurate to within approximately 2% when proper droplet trajectory con-
siderations are used in the data collection methods. The influence of droplet trajectory
and slit aperture alignment is investigated therein and found to be of great importance to
the in situ calculation of probe volume and subsequent volume flux calculation. For the
current investigations, a rotating slit aperture within the PDI receiver unit allowed for
alignment of the slit aperture with the local droplet trajectory and was adjusted for each
measurement point to acquire accurate volume flux measurements. The ability of the
PDI instrument to simultaneously measure drop size, velocity, and local flux provides an
attractive source of point measurements in a spray; however, the postprocessing meth-
ods outlined in this article are applicable to spray data acquired through any appropriate

z
y x

FIG. 5: PDI setup demonstrating the laser transmitter, receiver, and nozzle orientations.

Atomization and Sprays

Planar Statistics from Point Measurements 1015

point-sampling instrument or method, including computational results. Implementing

these methods in the postprocessing of computational results, which inherently have ac-
cess to these critical spray characteristics at a high spatial resolution through an entire
spray pattern, is attractive.

2.4 Weighted Average Methods

Weighted averaging is a reasonable and robust method for combining many individual
measurements into meaningful combined statistics, in this case, planar statistics. The
methods to generate weighted average values which are representative of the planar spray
plume characteristics follow typical mathematical processes; see, for example, the text
by Bevington and Robinson (2003). For the purposes of this article, which focuses on
spray measurements, the necessary practical steps are presented to generate weighted
averages based on the discrete measurement area, A, and local volume flux, qz .
For each parameter that is used for weighting, Pi,n , the relative weight,

Pi,n
ωPi,n = (1)
Σn Pi,n

is determined at each measurement point, where i is the weighting parameter index and
n is the measurement point index. When incorporating multiple weighting parameters,
as will be done here, a combined relative weight at each measurement location,
Σi ωPi,n
ωn = (2)
Σn Σi ωPi,n

must be determined. This combined normalized weight at each measurement location

using multiple parameters may then be implemented to generate a weighted average for
measured spray characteristics of interest. Note that with proper weighting across all
measurement points, the sum of all normalized ωn values will equal 1; this is also true
for any ωPi,n . The resulting weighted average value,

Xw = Σn (Xn ωn ) (3)

incorporates the relative distribution of the weighting parameters along with the mea-
sured quantity at each discrete location, Xn .
The total weighting parameter, ωn , and the measured values at each point will be
used to generate planar weighted average results in the xy plane here. A weighted aver-
age value, Xw , may be calculated for any point measurement value and will be calculated
in this article for the Sauter mean diameter, D32,w , and axial velocity, v z,w , in Section 4.
Two weighting parameters will be used in this investigation, and each represents a
relevant physical characteristics of the investigated spray plumes: (1) the discrete area

Volume 26, Issue 10, 2016

1016 Bade & Schick

that each measurement point is expected to represent and (2) the local volume flux at
the location of the measurement. The use of the two parameters together weights the
averaged results toward the measured values where the majority of the sprayed volume
is located. While these two parameters represent logical weighting values, other relevant
parameters could be used or added following the preceding methods.

2.4.1 Discrete Area Weighting

Discrete area weighting accounts for the fact that each measurement point is often not
expected to be representative of the same discrete area within a spray plume. An example
where this will have a large influence is in the implementation for results with a round
spray plume in which uniformly spaced radial measurements are acquired along a single
center-to-edge line (i.e., a single measurement axis, N = 1), as demonstrated by the
rings in Fig. 6(a). This discrete area geometry is referred to as nozzle-centered in this
investigation, as the central spray plume point is coincident with the nozzle axis. In
Fig. 6(a), the center point (point 1) is expected to represent the spray characteristics of a
very small circular area compared to, for example, the outer most point (point 4), which
represents a much larger area ring. By including additional radial measurement axes
(N > 1), the relative area of each discrete region is reduced but remains nonuniform, as
demonstrated by the ring segment areas at point 4 in Figs. 6(b)–6(c).
The relative area for a given discrete point,

An (rn + ∆r/2)2 − (rn − ∆r/2)2

A∗n = = (4)
A1 N (∆r/2)2

is straightforward for a round spray, where n is the steps number from the center point

y y y

x x x

4 4 4
3 3 3
1 2 2b 1 2 1 2
(a) (b) (c)
FIG. 6: Nozzle-centered area weighting: schematic example of the area represented by
a subset of discrete measurement locations (•) for a round spray with N measurement
axes for (a) N = 1, (b) N = 2, and (c) N = 4.

Atomization and Sprays

Planar Statistics from Point Measurements 1017

(n = 1,2,...,nmax ), N is the number of radial measurement axes from the spray center,
rn refers to the radius from the center of the spray plume to the nth measurement point,
and ∆r is the distance between measurement points. Each A∗n is normalized with the
area of the central spray plume point, A1 , and an equivalent ratio is also captured by
8(n − 1)/N , independent of the actual point distances. Naturally, the area of the center
point is always A∗1 = A1 /A1 = 1. The area ratios between points 1–4 are provided in
Table 1 for Figs. 6(a), 6(b), and 6(c), where the number of radial measurement axes
from the spray center are N = 1, 2, and 4, respectively.
Table 1 demonstrates the relative area ratio for the arrangements up to N = 4, as
demonstrated in Fig. 6; additional radial measurement axes would follow a similar pat-
tern using Eq. (4) with greater N values. All points at the same distance from the spray
center would receive the same area weight: that is, in Fig. 6(b), point 2b would receive
the same area weight as point 2. The benefit of using the identified A∗n term is the re-
moval of any need to calculate the actual discrete area for each measurement point due
to the systematic relative area resulting from uniformly spaced measurement locations.
In practice, the area weighting may be conducted using either An or A∗n , because each
weighting parameter is normalized in Eq. (1).
It is worth noting that the idealized round spray patterns of Fig. 6 can also be con-
verted to oval patterns and that the relative area of each subsequent ring follows the same
area proportions. Thus, with a standard flat-spray pattern, as that given in Fig. 2 for the
flat spray investigated here, a nominally oval spray pattern is assumed and provides a
close representation of the relative area expected to be represented by radial subsets of
measurement points. The influence of specific radial measurement axes arrangements is
explored in the Section 4.
For the multiorifice spray, additional considerations are appropriate to properly
weight the more complex spray plume shape. Investigating the spray plume generated
by one of the six orifices, the plume shape is interpreted as triangles toward the noz-
zle center and as semicircles toward the outer edge of the spray relative to the y ′ -axis.

TABLE 1: Relative area to the center point area,

A∗n = An /A1 , of each subsequent measurement
location, n, for various numbers of radial mea-
surement lines from the spray center, N
n N =1 N =2 N =4
1 1 1 1
2 8 4 2
3 16 8 4
4 24 12 6
.. .. .. ..
. . . .

Volume 26, Issue 10, 2016

1018 Bade & Schick

For this nonaxisymmetric spray plume shape, determining a spray center point can be
subjective, and the nozzle center provides the only tractable starting point for regularly
spaced measurements. Therefore a method is devised to use uniformly spaced measure-
ment points to investigate a single plume of the multiorifice spray with measurements
beginning at the nozzle center (x = y = 0). Figure 7(b) provides two reasonable examples
of subsets of measurement points that may be used to generate the planar spray statistics
and the expected representative discrete areas of of each measured point value.
To determine the discrete areas shown in Fig. 7, the location of maximum volume
flux along the x-axis is taken as the dividing point (location of the y ′ -axis) for area
weighting, with the nozzle center located at point 1. This alternative geometry is referred
to herein as plume-centered weighting. This, and other, subsets of the full matrix of PDI
measurement points are used to generate combined planar spray statistics, to determine
a minimized subset of measurement points to collect accurate overall spray statistics for
this spray, in Section 4. A slightly different set of relative areas for each triangularly
shaped area region is required, which are 1, 5, 9, 13, and soon, starting at the y ′ -axis.

2.4.2 Volume Flux Weighting

Volume flux weighting accounts for regions within the spray plume that have a relatively
high or low flux to be weighted accordingly. For example, despite the center of a round
spray plume having a small discrete area that it represents relative to the spray plume
edge, the volume flux through the center location may be much higher than at the spray
edges. The PDI allows accurate simultaneous measurement of drop size, velocity, and
volume flux and therefore allows this weighting to be conducted efficiently without addi-
tional measurements. However, as a caution to investigators, in practice, accurate mea-
surement of volume flux involves detailed real time assessment of the effective probe
volume and, without careful and often tedious provisions to account for errors due to

y y’ y y’

x x

1 1
2 2
3 3
4 4
(a) (b)

FIG. 7: Plume-centered area weighting: schematic example of the area represented by a

subset of discrete measurement locations (•) for a teardrop-shaped spray generated by a
multiorifice nozzle with N number of measurement axes for (a) N = 2 and (b) N = 4.

Atomization and Sprays

Planar Statistics from Point Measurements 1019

coincidence (Bachalo et al., 1988), slit aperture (Albrecht et al., 2003), droplet trajec-
tory (Albrecht et al., 2003), Gaussian beam effects/probe volume correction (Albrecht et
al., 2003), and combined droplet trajectory/slit aperture effects (Bade and Schick, 2011),
can result in inaccurate volume flux measurement (by an order of magnitude; Bade and
Schick (2011)) and thus unreliable values for weighting. The methods used to obtain
accurate volume flux measurements in the present study follow those described by Bade
and Schick (2011).

3. POINT MEASUREMENT RESULTS

The results of the PDI pointwise measurements for each nozzle are presented in this
section; for brevity, these results focus on the Sauter mean diameter, D32 , and axial
velocity, vz ; however, the evaluations are representative of other spray characteristics
(such as D10 or Dv0.5 ).
For the round spray, a total of 43 measurement points were collected with the PDI
along the x-axis, as demonstrated in Fig. 1. Figure 8 provides the profile results, at
z = −100 mm from the nozzle, for the 1/8G-1 nozzle spray. The results demonstrate the
fairly axisymmetric distribution of this full-cone nozzle; a swirl element located within

300 12
x (at y=0) x (at y=0)
250 10
v z (m/s)
D 32 (µm)

200 8
150 6
100 4
50 2
0 0
−80 −60 −40 −20 0 20 40 60 80 −80 −60 −40 −20 0 20 40 60 80
x (mm) x (mm)
(a) (b)
0.4
x (at y=0)
q z (cm 3 /cm 2 /s)

0.3

0.2

0.1

0
−80 −60 −40 −20 0 20 40 60 80
x (mm)
(c)
FIG. 8: PDI point measurement results for the 1/8G-1 single-orifice, hydraulic, round
spray for (a) D32 , (b) vz , and (c) qz .

Volume 26, Issue 10, 2016

1020 Bade & Schick

the nozzle just upstream of the exit does result in subtle nonaxisymmetry that be seen
in the LSI results of Fig. 1; and this asymmetry is reduced at larger spray distances.
The drop size is shown to vary by approximately 25% over the spray pattern, and, as
expected, both the velocity and volume flux demonstrate a maximum at the spray cen-
ter and decrease at larger radial locations. Integration of the local volume flux mea-
surements, over the appropriate representative discrete areas [i.e., as demonstrated in
Fig. 6(b)], yields a total volume flow rate of Qz = 728 mL/min, where the measured total
flow rate was Qtotal = 710 mL/min – an error of 2.5%, which is quite accurate and lends
confidence to the volume flux measurements that will be used for weighted averaging
purposes.
For the flat spray, a total of 273 measurement points were collected with the PDI
according to the matrix demonstrated in Fig. 2. Figure 9 provides the resulting xy-planar
results, at z = −50 mm from the nozzle, for the TPU650050 nozzle spray. Figure 10
provides the extracted x- and y-axes profile results, which will be used extensively in
the reduced data set analysis of Section 4, but also demonstrates the strong symmetry
about both the x- and y-axes, while not being axisymmetric.
The results for D32 and vz in Fig. 9 demonstrate that these quantities are nonaxisym-
metric but are fairly symmetric about the x- and y-axes. Therefore weighted averages

30 30 vz
D 32
20 (µm) 20 (m/s)
200 15
y (mm)

10
y (mm)

10
0 0 10
100
−10 −10 5
−20 −20
0 0
−30 −30
−50 −40 −30 −20 −10 0 10 20 30 40 50 −50 −40 −30 −20 −10 0 10 20 30 40 50
x (mm) x (mm)
(a) (b)
30 qz
20 (cm 3 /cm 2 /s)
1
y (mm)

10
0
0.5
−10
−20
0
−30
−50 −40 −30 −20 −10 0 10 20 30 40 50
x (mm)
(c)
FIG. 9: PDI point measurement results for the TPU650050 single-orifice, hydraulic, flat
spray for (a) D32 , (b) vz , and (c) qz .

Atomization and Sprays

Planar Statistics from Point Measurements 1021

250 20
x (at y=0) x (at y=0)
200 y (at x=0) y (at x=0)
15
D 32 (µm)

v z (m/s)
150
10
100
5
50

0 0
−50 −40 −30 −20 −10 0 10 20 30 40 50 −50 −40 −30 −20 −10 0 10 20 30 40 50
x or y (mm) x or y (mm)
(a) (b)
1.5
x (at y=0)
q z (cm 3 /cm 2 /s)
1.2 y (at x=0)

0.9

0.6

0.3

0
−50 −40 −30 −20 −10 0 10 20 30 40 50
x or y (mm)
(c)
FIG. 10: PDI point measurement results across the x- and y-axes for the TPU650050
single-orifice, hydraulic, flat spray for (a) D32 , (b) vz , and (c) qz .

based on the volume flux distribution, qz , will likely require multiple measurement axes
from the center point. Furthermore, these results clearly show that measurements along
the x-axis (at y = 0) will not adequately capture the spray characteristics. This is instruc-
tive, because the nominally oval spray pattern may suggest that the spray characteristics
would be fairly axisymmetric, an incorrect assumption that is strengthened by testing
results like those of the LSI surface area distribution of Fig. 2. It is clear from Figs. 9(a)
and 9(b) that the D32 and vz planar distributions exhibit increased values as the distance
from the spray center is increased in x (at y = 0). In contrast, D32 and vz undergo a
monotonic decrease as the distance from the center is increased in y (at x = 0). The re-
sults of the LSI testing are interesting in light of the PDI results, where in Fig. 9(c) the
largest concentration of spray material is coincident with the regions of larger velocity
and drop size (larger x regions where y = 0). However, the LSI results (which are rel-
ative to the local spray surface area) indicate the largest values at the spray center. For
these two distribution results to both be accurate, the drop counts must be much larger
in the spray center (where the droplets are smaller). This conclusion is confirmed by the
PDI data acquisition rate, which shows a similar relative distribution as the LSI result,
with the largest frequency of droplet passage through the measurement plane at the spray
center and a monotonic decrease as x and y are increased. Integration of local volume

Volume 26, Issue 10, 2016

1022 Bade & Schick

flux measurements over the appropriate representative areas (i.e., the discrete measure-
ment grid area) yields a total volume flow rate of Qz = 238 mL/min, while the measured
flow rate was Qtotal = 225 mL/min – an error of 5.6%, which again lends confidence to
the PDI measurements. Note that the authors have observed many research papers that
uniformly scale the local volume flux measurements obtained from phase Doppler in-
struments to match the integrated and actual total flow rate; this was not done here and
is generally a poor technique. As demonstrated by Bade and Schick (2011), inaccurate
volume flux measurements are typically not uniformly inaccurate.
For the multiorifice spray, a total of 453 measurement points were collected with the
PDI according to the layout demonstrated in Fig. 3, covering one of the six spray plumes
generated by the nozzle. Figure 11 provides the xy planar results, at z = −178 mm,
from these tests with the SU26B nozzle spray. Figure 12 provides the extracted x- and
y-axes profile results, which will be used extensively in the reduced data set analysis of
Section 4, and demonstrates the strong symmetry about the y-axis but not the x-axis.
In Fig. 11(a), the Sauter mean diameter is shown to increase near the spray edges.
This is typical of an air-atomized spray where the core of the spray plume tends to exhibit
only small variations in drop size, while at the edges the liquid is influenced by the nozzle
exit orifice walls and larger droplets can be generated. Furthermore, the small droplets

50 50 vz
D 32 40
40 (µm) (m/s)
30 80 30 8
20 20
y (mm)
y (mm)

10 60 10 6
0 0
−10 40 −10 4
−20 −20
20 −30 2
−30
−40 0
−40 0
−50 −50
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
x (mm) x (mm)
(a) (b)
50 qz
40 (cm 3 /cm 2 /s)
30
20
y (mm)

0.01
10
0
−10 0.005
−20
−30
−40 0
−50
0 20 40 60 80 100 120 140
x (mm)
(c)
FIG. 11: PDI point measurement results for the SU26B multi-orifice, air-atomizing
spray for (a) D32 , (b) vz , and (c) qz .

Atomization and Sprays

Planar Statistics from Point Measurements 1023

100
12
x (at y=0)
x (at y=0)
y (at y ′ ) 10 y (at y ′ )
80

v z (m/s)
D 32 (µm)
8
60
6
40 4
20 2
0 0
−50 −25 0 25 50 75 100 125 150 −50 −25 0 25 50 75 100 125 150
x or y (mm) x or y (mm)
(a) (b)
0.025
x (at y=0)
q z (cm 3 /cm 2 /s)
y (at y ′ )
0.02

0.015

0.01

0.005

0
−50 −25 0 25 50 75 100 125 150
x or y (mm)
(c)
FIG. 12: PDI point measurement results across the x- and y-axes for the SU26B multi-
orifice, air-atomizing spray for (a) D32 , (b) vz , and (c) qz .

are easily carried by the airstream that exits the nozzle, while larger droplets may cross
streamlines and move toward larger radial locations. As expected, the velocity of the
droplets tends to decrease near the spray edges as the atomizing air forms a jet that
dissipates outward from the center of the plume. Because the SU26B is a multiorifice
nozzle, each exit orifice is set to an angle relative to the measurement plane, and the
atomizing air that carries the droplets has been further slowed by the longer path to the
larger x locations. Examining Fig. 11(c), the x location of largest volume flux, at y = 0, is
selected as the plume center point for area-weighting analysis, as demonstrated in Fig. 7.
The plume center point is found to be at approximately (x, y) = (90, 0) mm here, and the
y ′ -axis is defined here. This point correlates well with the velocity peak, which occurs
at nominally the same location. Integration of local volume flux measurements over the
appropriate discrete areas yields a total volume flow rate of Qz = 147 mL/min, where
the measured total flow rate was Qtotal = 136 mL/min – an error of 8.1%.
In the results provided in Figs. 1–3, the PDI measurement arrays can be seen to
extend outside of the final contour line, which is indicative of <5% local surface area
density of the spray. This use of the LSI results to determine the spray plume limits
worked well, as the PDI results for volume flux reached a negligible level at the mea-
sured spray edges. However, the results for volume flux distribution and surface area

Volume 26, Issue 10, 2016

1024 Bade & Schick

distribution clearly demonstrate differences; in particular, the center-to-edge distribution

for the flat spray (Fig. 2 vs. Fig. 9) is quite notably different. These differences between
the LSI and PDI volume flux contours is due to a change in the width of the drop size
distributions at different locations within each spray. The hydraulic-flat spray has very
different drop size distribution widths at the center versus edge locations, and therefore
the surface area versus volume flux contour gives a different result. The internally air-
atomized multiorifice spray has a similarly wide drop size distribution over much of the
spray pattern, resulting in similar relative distributions of spray surface area and spray
volume flux.

4. PLANAR STATISTICS RESULTS

An investigation into the measurement point locations necessary to calculate accurate

planar spray statistics as well as the necessary weighted averaging (volume flux and
area) methods are analyzed here. As described in Section 2.4, many subsets of the PDI
data were used to generate mean statistical values with the straight mean values denoted
with an m subscript (e.g., xm represents points along the x-axis used in a straight average
calculation), weighted average values use a w subscript (e.g., xw represents points along
the x-axis used in a weighted average), and half-axis weighted average values (e.g., +xw
represents a weighted average results using points along only one-half of the x-axis).
The resulting straight and weighted average planar values are presented in Figs. 13–15
when performing measurements along the x-, y-, and x- and y-axes, along with the error
for each calculated value on the inset axes. Smaller values of the number of sampling
points, n, used in the weighted averaging were achieved by down-sampling the number
of measurement points (e.g., using every other point) used along each axis to simulate
a reduced set of measurement points. Finally, a baseline measurement was performed
by using the full grid of measurements, which was down-sampled as well; the correct

250 12
6 6 baseline
baseline xm
e vz (%)
e D3 2 (%)

xm 4 xw
240 4 xw 11
+x w +x w
2 2
v z (m/s)

230
D32 (µm)

10 0 0
0 0 1 2 3 1 2 3
10 10 10 10 10 10 10 10
220 n n
9
210
8
200

190 7
0 10 20 30 40 50 0 10 20 30 40 50
n (number of points) n (number of points)

(a) (b)
FIG. 13: Weighted average planar values using various subsets of data along the x-axis
for the 1/8G-1 round spray for (a) D32 and (b) vz . The inset figure provides the error
using each subset.

Atomization and Sprays

Planar Statistics from Point Measurements 1025

200 24
25 baseline 25 baseline
grid 22 20 grid
20

e vz (%)
e D3 2 (%)
xm xm
180 15 20 15 ym
ym
10 xy m 10 xy m

v z (m/s)
D32 (µm)
xw 18 xw
160 5 yw 5 yw
0 0 xy w 16 0 0 1 2 3
xy w
1 2 3
10 10 10 10 +x w 10 10 10 10 +x w
140 n +y w 14 n +y w
+xy w +xy w
12
120
10
100 8
0 50 100 150 200 250 300 0 50 100 150 200 250 300
n (number of points) n (number of points)

(a) (b)

FIG. 14: Weighted average planar values using various subsets of data along the x- and
y-axes for the TPU650050 flat spray for (a) D32 and (b) vz . The inset figure provides the
error using each subset.

9
40
30
30
baseline baseline
25 25 grid
grid 8

e vz (%)
e D3 2 (%)

38 20 xm
20 xm
ym 15 ym
15 v z (m/s) 7 xy m
36 xy m
D32 (µm)

10 10 xw
xw
5 yw 5 yw
34 6 xy w
0 0
xy w 0 0
1 2 3 1 2 3
10 10 10 10 10 10 10 10
32 n 5 n

30 4

28 3
0 100 200 300 400 500 0 100 200 300 400 500
n (number of points) n (number of points)

(a) (b)

FIG. 15: Weighted average planar values using various subsets of data along the x- and
y-axes for the SU26B multiorifice spray for (a) D32 and (b) vz . The inset figure provides
the error using each subset.

measurement is taken as the value generated by the volume flux and area weighted values
calculated with the full grid of measurement points for each spray (and therefore the
highest number of points) and is noted on each plot as the baseline. The results of smaller
subsets of data are evaluated for error relative to the baseline value.
In Fig. 13, the round spray planar statistics are evaluated and the error in the cal-
culated D32 and vz is presented for the various subsets and number of points. Note
that the baseline and the x-axis measurements are the same in this case, because mea-
surements were only acquired along the x-axis in this nominally axisymmetric spray.
Clearly the accuracy increases as the number of points grows, but notably, the error from
all weighted average subsets is less than 4%, with the majority of error falling below 2%.
The straight average results also show good agreement with the baseline measurement
and have less than 5% error. For this axisymmetric spray, it is therefore recommended

Volume 26, Issue 10, 2016

1026 Bade & Schick

to measure approximately eight points from the center to edge of the spray to calculate
a planar measurement within 2% of the actual value – or roughly one point every 10%
of the spray plume. Alternatively, measuring 11 points spanning the entire spray plume
from edge to edge results in an error of less than 1% – an equivalent spacing of one
measurement for every 10% of the overall spray plume width.
The flat spray planar statistics are presented in Fig. 14. Notably, for measurements
made across only the x-axis of the flat spray plume, the D32 weighted average is over-
estimated by approximately 23% and the vz is overestimated by approximately 16%,
independent of the number of measurement points. This is a noteworthy result given that
flat sprays are typically characterized along the major axis (x-axis). Examining Fig. 9(a),
the erroneously large D32 result is due to the large drop size regions near the flat spray
edges along the x-axis; similarly, in Fig. 9(b), the largest velocities are centered around
the x-axis and thus provide an inappropriate representation of the total spray plume
characteristics. For subsets including only measurements along the y-axis, large errors
in average D32 and vz (approximately 10% and 15%, respectively) are also found due to
measurements not providing a representative account of the spray characteristics along
the x-axis edges, regardless of the number of measurement points. Finally, measure-
ments performed along both the x- and y-axes resulted in much higher accuracy for both
drop size and axial velocity. It is shown that errors as low as 5.1% in D32 and 1.8% in
vz are achieved using subsets of only 23 measurement points, which is a significant re-
duction from the full data set array of 273 points. The straight average results mimic the
weighted average accuracy; this is assumed to be fairly coincidental, and the weighted
average results should provide a more robust and representative result. From a practical
standpoint, measurements that span half of both the x- and y-axes (i.e., center to edge of
the spray pattern in each direction) are able to capture the D32 and vz within 5% of the
full grid value, using a total of eight measurements (5 in x and 4 in y with a common
measurement at the spray center point) – equivalent of one measurement point for every
20% of the spray plume height and width.
For the multiorifice nozzle results (Fig. 15), the atomization mechanism are changed
with the presence of atomizing air, which is introduced within the nozzle body (inter-
nally). The resulting spray plume has larger droplets near the plume edge, and the peak
velocity and volume flux occur near the plume center. Regarding the planar drop size
results, there is no significant difference found by taking measurements along the x-,
y-, or x- and y-axes; the number of points collected along each axis also played an in-
significant role. Recall that the y-axis measurements are acquired along the y ′ -axis here,
defined at the x-location of highest volume flux at y = 0. The resulting planar average
D32 values all fall within 4% to 9% of the baseline value. These results show that there
is no appreciable benefit in planar drop size statistics measurement from including mea-
surement points off the x-axis, so long as the weighting is conducted so that the center
point is taken as the location of largest volume flux, as described in the text surrounding
Fig. 7. Furthermore, the straight averaged results actually outperform the weighted aver-

Atomization and Sprays

Planar Statistics from Point Measurements 1027

age using measurements along the x-axis. This is likely due to the internally air-atomized
nature of the SU26B nozzle, which produces a fairly uniform drop size throughout the
spray plume (as is discussed around Fig. 16(b). Therefore, to achieve the best accuracy
using measurements along only the x-axis, a minimum of 12 points should be collected
representing a measurement nearly each 10% of the total spray plume size, and a straight
average can be used.
However, the weighted average planar axial velocity values, v z , demonstrate a sig-
nificant improvement when measuring along both the x- and y-axes, reaching errors of
1% to 4%. Examining Fig. 11(b), measurements along the x-axis only result in an over-
weighting of the low velocity region, 0 < x < 60 mm, and an error of approximately
20%. Conversely, measurements along only the y-axis do not capture the low velocity
in the 0 < x < 60 mm region and result in overestimation errors of around 20%. There-
fore, for accurate planar velocity results, measurements must be acquired along with the
x- and y ′ -axes, and acquiring measurements at every 20% of the total spray plume size
results in errors of less than 5%. Thus, in summary, to acquire planar drop size and ve-
locity statistics for a nozzle similar to the multiorifice SU26B nozzle investigated here,
measurements would need to be acquired at each 10% of the overall spray plume in both
the x and y directions.
Finally, with a sufficient evaluation of the accuracy of combined point measurement
mean values, the resulting conclusions may be applied to combining droplet size distri-
butions collected at many discrete locations. Normalized drop size count distributions
from each point measurement may be weighted and combined bin-wise to generate pla-
nar drop size distributions using the local volume flux and discrete area weighted aver-
aging methods. Figure 16 demonstrates the individual drop size distributions at all mea-

10 16
9 Local Local
Mean 14
8 Mean
Weighted Mean 12
Counts (%)

Counts (%)

7 Weighted Mean
6 10
5 8
4 6
3
4
2
1 2
0 0
0 50 100 150 200 250 300 350 400 0 20 40 60 80 100 120
Diameter (µm) (2 µm bins) Diameter (µm) (2 µm bins)
(a) Flat spray drop size distributions (b) Multiorifice spray drop size distributions

FIG. 16: Drop size distributions at the (a) 273 individual PDI measurement points ac-
quired with the TPU650050 flat spray nozzle and the (b) 453 individual PDI measure-
ment points acquired with the SU26B multiorifice nozzle, along with the straight average
and weighted average drop size distributions.

Volume 26, Issue 10, 2016

1028 Bade & Schick

surement locations for the flat and multiorifice sprays, along with the straight average
and weighted average distributions (using all points, but could also be performed using
a reduced set of test points). These holistic representations of the spray plume drop size
distributions provide a unique and new method to combine and represent global spray
characteristics. In Fig. 16(a), the wide range of drop size distributions for the hydraulic
flat spray can be seen, resulting in straight and weighted average distributions that are
not the same. Alternatively, the internally atomized multiorifice spray individual dis-
tributions, shown in Fig. 16(b), all demonstrate a consistent shape, resulting in nearly
identical average and weighted average distributions. The differences between the the
atomization mechanisms of the hydraulic flat spray and internally air-atomized multior-
ifice spray result in a much more uniform drop size and drop size distribution across the
air-atomized spray.

5. CONCLUSIONS

This study presents methods to generate planar mean spray statistics from small sets of
point measurements using discrete area and volume flux weighting. The results demon-
strate that the developed methods produce accurate results and may result in significant
testing-time savings by measuring at a small number of point locations with a spray.
The detailed results of this study are nozzle and operating condition specific; however,
the volume flux and discrete area weighting methods are applicable to many spray mea-
surement efforts and allow robust, holistic comparisons between nozzles and operating
conditions. Furthermore, the evaluation of the effect of many different practical axes
over which the individual measurement points may be collected provides direction for
future investigators. It is found that for a fairly axisymmetric, hydraulic, round spray, a
single axis measurement program is sufficient to capture the spray statistics, with mea-
surements at each 10% of the overall spray plume from edge to edge through the spray
center. For a hydraulic flat spray, it is necessary to acquire measurements along both the
major and minor axes (x and y) to generate accurate planar mean values. Finally, it is de-
termined for an internally air-atomized, multiorifice spray plume that it is only necessary
to collect measurements from the nozzle center to the plume edge (x-axis here), through
the center of one of the plumes, to generate accurate planar drop size values. However,
to generate accurate velocity values, it is also necessary to collect measurements along
the y-axis, and it is recommended that these measurement be aligned with the volume
flux peak along the x-axis. All the measurement recommendations result in planar mean
values for drop size and velocity within approximately 5% of the actual value. Finally,
combined drop size distributions are demonstrated for both the hydraulic flat spray and
the multiorifice internally air-atomized spray and highlight the notably different droplet
size distributions throughout each spray type.

Atomization and Sprays

Planar Statistics from Point Measurements 1029

ACKNOWLEDGMENTS

The authors would like to acknowledge and thank Krunal Patel and Ben Bridges of
Spraying Systems Co. for their efforts in the laboratory in support of this work.

REFERENCES

Albrecht, H.-E., Borys, M., Damaschke, N., and Tropea, C., Laser Doppler and Phase Doppler
Measurement Techniques, Berlin: Springer, 2003.
Bachalo, W. D., Rudoff, R. C., and Brena de la Rosa, A., Mass flux measurements of a high
number density spray system using the phase Doppler particle analyzer, AIAA 26th Aerosp.
Sci. Meeting, Reno, NV, 1988.
Bade, K. M. and Schick, R. J., Volume distribution comparison methods for 1D, 2D, and point
measurement techniques, ILASS-Americas 21st Annual Conf. on Liquid Atomization and
Spray Systems, Orlando, FL, 2008.
Bade, K. M. and Schick, R. J., Phase doppler interferometry volume flux sensitivity to parametric
settings and droplet trajectory, Atomization and Sprays, vol. 21, no. 7, pp. 537–551, 2011.
Bade, K. M. and Schick, R. J., Post-processing of phase doppler interferometry data for planar
spray characteristics, ILASS-Americas 25th Annual Conf. on Liquid Atomization and Spray
Systems, Pittsburgh, PA, 2013.
Bevington, P. R. and Robinson, D. K., Data Reduction and Error Analysis for the Physical Sci-
ences, 3rd ed., New York: McGraw-Hill, 2003.
Chapple, A. C. and Hall, F. R., A description of the droplet spectra produced by a flat-fan nozzle,
Atomization and Sprays, vol. 3, no. 4, pp. 477–488, 1993.
Dodge, L. G., Rhodes, D. J., and Reitz, R. D., Drop-size measurement techniques for sprays:
Comparison of malvern laser-diffraction and aerometrics phase/doppler, Appl. Opt., vol. 26,
no. 11, pp. 2144–2154, 1987.
Falser, T. D. and Parrish, S. E., Spray measurement technology: A review, Measure. Sci. Technol.,
vol. 26, pp. 1–34, 2015.
Gandhi, A. H., Shakal, J. S., Markle, L. E., Hung, D. L. S., Cummings, S. D., Parrish, S. E., Har-
rington, D. L., Sayar, H., Kramer, J. L., and Meinhart, M. A., A new measurement standard for
the characterization of automotive fuel sprays, ILASS-Americas 21st Annual Conf. in Liquid
Atomization and Spray Systems, Orlando, FL, 2008.
Hung, D. L. S., Harrington, D. L., Gandhi, A. H., Markle, L. E., Parrish, S. E., Shakal, J. S.,
Sayar, H., Cummings, S. D., and Kramer, J. L., Gasoline fuel injector spray measurement and
characterization – A new SAE J2715 recommended practice, SAE Technical Paper 2008-01-
1068, 2008.
Lefebvre, A. H., Atomization and Sprays, Combustion: An International Series, Singapore:
Hemisphere, 1989.
Mugele, R. A. and Evans, H. D., Droplet size distributions of sprays, Ind. Eng. Chem., vol. 43,
no. 6, pp. 1317–1324, 1951.

Volume 26, Issue 10, 2016

1030 Bade & Schick

SAE International, Gasoline Fuel Injector Spray Measurement and Characterization, SAE Stan-
dards Document J2715, Warrendale, PA: SAE International, 2007.
Tishkoff, J. M., Ingebo, R. D., and Kennedy, J. B. (eds.), Liquid Particle Size Measurement
Techniques, special technical publication 848, American Society for Testing and Materials
(ASTM), 1984.
Tuck, C. R., Butler, M. C., and Miller, P. C. H., Techniques for measurement of droplet size
and velocity distributions in agricultural sprays, Crop Protection, vol. 16, no. 7, pp. 619–628,
1997.

Atomization and Sprays