You are on page 1of 11

ANALYZING THE NOZZLE SPRAY FAN PATTERN OF AN

AGRICULTURAL SPRAYER USING PULSE WIDTH


MODULATION TECHNOLOGY TO GENERATE
AN ON-GROUND COVERAGE MAP
D. L. Mangus, A. Sharda, A. Engelhardt, D. Flippo, R. Strasser, J. D. Luck, T. Griffin

ABSTRACT. Chemical application is an integral part of crop care. Today, advanced sprayers automatically control indi-
vidual boom sections and nozzles to accommodate increased machine sizes and travel speeds, yet automatic control of
flow-based systems raises concerns regarding coverage accuracy and uniformity during changes in travel speed and
spray swath width. New commercial systems apply product at a constant pressure using varied duty cycles of pulse width
modulated (PWM) solenoids to maintain a constant application rate. However, concerns exist regarding the dynamic ef-
fect of solenoid on/off latency on spray fan pattern and spray coverage. The objectives of this study were to investigate the
on/off latency in PWM nozzles, determine if active nozzles affect spray fan pattern latency, and develop flow characteris-
tics to simulate dynamic spray coverage. A PWM system and flow rate controller were installed on a 6.6 m three-section
boom sprayer with 13 nozzles. A Raven Viper 4 controller regulated the product flow rate and pressure, while a Capstan
Pinpoint controller was used to set the system pressure, nozzle on/off configuration, and duty cycle. The results indicated
that the PWM spray system maintained the pressure within 5% of the target value and applied an accurate amount of
flow per pulse regardless of the number of nozzles activated. There was a 20 ms delay in nozzle pressure development
during each cycle, and the delay was constant regardless of the number of nozzles activated. After de-energizing the sole-
noid, the nozzle continued spraying at system pressure for 10 ms. Static spray droplet distribution proved that the system
applied the correct volume per pulse. In addition, PWM duty cycles of 100%, 80%, 60%, and 40% provided spray cover-
age within 10% of the target rate for 100%, 94%, 77%, and 67% of the time, respectively. Greater signal overlap be-
tween odd and even nozzles increased the application coverage. Dynamic spray simulations showed that as-applied appli-
cation error may vary beyond 10% of the target rate. As such, while the PWM system provided the desired amount of
product per pulse, the spray coverage results indicated that the on-ground coverage could result in areas with under- or
over-application.
Keywords. Dynamic spray coverage simulation, High-speed imagery, Individual nozzle, Precision ag, Pulse width modu-
lation, Spray fan pattern.

T
oday, many farmers are transitioning from con- applications to kill pests, such as weeds and insects, to keep
ventional tillage to conservation and no-till farm- organic matter on the soil surface, thereby preventing wind
ing practices, leading to fewer field passes, in- and water erosion, and to decrease soil degradation. Gener-
creased organic matter, and minimum soil com- ally, chemical application costs can exceed one-third of the
paction, which have led to reduced input costs and in- total cost of crop production (Grisso et al., 1989). In 2010,
creased crop yields. Conservation tillage uses chemical U.S. farmers spent a total of $11.5 billion on pesticides
alone. To account for chemical costs, increased machine
size and control capabilities have been shown to increase
the land’s productivity, improve application efficiency,
Submitted for review in March 2016 as manuscript number MS 11835; sustain production growth, and increase environmental
approved for publication by the Machinery Systems Community of
ASABE in January 2017. Presented at the 2015 ASABE Annual Meeting stewardship (Sharda et al., 2010, 2011). With increased
as Paper No. 152189633. machine size and travel speed, irregular-shaped fields cre-
The authors are Devin L. Mangus, ASABE Member, Graduate ate concerns because of the chance of skips, over-
Student, Ajay Sharda, ASABE Member, Assistant Professor, Andrew
Engelhardt, Engineer, Daniel Flippo, ASABE Member, Assistant
application, and unintentional application (Porter et al.,
Professor, and Ryan Strasser, ASABE Member, Graduate Student, 2013; Sama et al., 2015). Such errors raise concerns about
Department of Biological and Agricultural Engineering, Kansas State environmental pollution, erosion of sensitive areas such as
University, Manhattan, Kansas; Joe Luck, ASABE Member, Assistant grassed waterways, and overuse of fertilizers and pesti-
Professor, Department of Biological Systems Engineering, University of
Nebraska, Lincoln, Nebraska; Terry Griffin, ASABE Member, Assistant cides, which contribute to increased production costs. Ir-
Professor, Department of Agricultural Economics, Kansas State regular-shaped fields also create the need for automatic
University, Manhattan, Kansas. Corresponding author: Ajay Sharda, section control (ASC) to turn individual nozzles and boom
145 Seaton Hall, Kansas State University, Manhattan, KS 66506; phone: sections on and off to maximize coverage, minimize over-
785-532-2936; e-mail: asharda@ksu.edu.

Transactions of the ASABE


Vol. 60(2): 315-325 © 2017 American Society of Agricultural and Biological Engineers ISSN 2151-0032 DOI 10.13031/trans.11835 315
lap, and decrease off-target application, especially on sur- study was designed with the following objectives: (1) quan-
face water sources (Porter et al., 2013). tify spray fan on/off time latency for a spray system using
Errors in chemical application are classified as static or PWM technology, (2) determine if the number of active
dynamic. Static errors result from incorrect machine setup nozzles affects nozzle pressure and spray fan pattern laten-
(i.e., nozzle type and size, calibration, and system checks), cy, and (3) generate simulated on-ground dynamic spray
chemical mix concentrations, and operating tolerances for coverage using empirical spray droplet distribution charac-
section control. Dynamic errors result from overlapping teristics.
coverage, changes in travel speed, and changes in boom
height over elevation changes in the terrain (Porter et al.,
2013). Automatic control has enabled sprayers to increase METHODS AND MATERIALS
coverage area and overall machine size, maintain desired A commercial control system for individual nozzles was
pass-to pass distance and travel speeds, and decrease over- evaluated in the Department of Biological and Agricultural
lap. In highly irregular-shaped fields, error reductions of up Engineering at Kansas State University. The PWM system
to 17% were noted in comparing 30-section automatic noz- used inline solenoids to provide automatic proportional
zle control with boom section control (Luck et al., 2010). flow at each nozzle based on the travel speed of the sprayer
Individual nozzle control could provide precise chemical (km h-1) and target application rate (L ha-1). Using a digital
control, decreasing the permitted minimal management source, the PWM system generates an analog signal that
zone (Luck et al., 2011a). has two key components: duty cycle and frequency. The
Large self-propelled sprayers typically use flow-based duty cycle represents the amount of time that the signal is
controllers that maintain the application rate by matching in a high (on) state as a percentage of the total time to com-
the product flow rate to the travel speed and spray swath plete one cycle. The frequency determines how fast the
during ASC actuation. Past research indicated that nozzle PWM system completes a cycle, or how fast it switches
pressure deviations occurred during ASC actuation (Sharda between high and low states. The frequency (in Hz) repre-
et al., 2010), and spray application with self-propelled sents the cycles per second. All commercially available
sprayers was accurate for only 35% of the time during field PWM systems from manufacturers such as Capstan, Raven,
applications (Sharda et al., 2011). Additional studies con- and Teejet use 10 cycles per second, or a 100 ms cycle, to
ducted on these systems to quantify controller response operate the solenoids mounted on each nozzle. For this
during turning movements indicated that the application study, a Pinpoint system (Capstan Ag Systems, Inc., Tope-
rate deviated considerably from the target rate for turning ka, Kans.) and Wilger combo-jets nozzles (model MR110-
radii of up to 30 m (Luck et al., 2011b). The application 05, Wilger Inc., Lexington, Tenn.) were selected.
errors were typically attributed to response delays in the
flow control hardware that prevented timely implementa- PARTICLE AND SPRAY FAN ANALYSIS
tion of flow rate changes in the plumbing system (Sama et The size distribution and droplet velocity from a Wilger
al., 2015). The magnitude of the application errors was also MR 110-05 nozzle were measured using a VisiSize sizing
related to the number of boom sections and individual noz- system (Oxford Lasers, Inc., Shirley, Md.) (fig. 1). From
zle control sections used during ASC actuation (Sharda et these data, the droplet size, velocity, and volume distribu-
al., 2011). Typically, over-application occurred when ASC tion were quantified to create the flow characteristics of a
actuation turned sections off and during sprayer decelera- single nozzle when operated at 345 kPa with a spray boom
tions, and under-application occurred when ASC actuation height of 53.3 cm. Droplet speeds were grouped into six
turned sections on and during sprayer accelerations (Sharda speed intervals, and the droplet size range and cumulative
et al., 2012). Past studies also indicated that nozzle pressure percent volume for each speed interval was computed
increased or decreased with similar demands in product (table 1). In practice, some droplets will not follow the re-
flow rate in the plumbing system during ASC actuation and sults shown in table 1; the speed distributions in table 1
during speed transitions (Sharda et al., 2011). Increased represent a uniform spray distribution obtained from three
nozzle pressure typically changes the droplet size distribu- replicated tests for simulation purposes, as described later.
tion and spray fan angle (Teejet, 2014), which could poten-
tially impact the overlap of adjacent nozzles and the cover-
age accuracy.
As an alternative to flow-based controllers, newer self-
propelled sprayers are equipped with pulse width modula-
tion (PWM) technology. PWM systems use solenoid-
operated nozzles to control the product flow rate from each
nozzle (Sama et al., 2015). The PWM system is pro-
grammed for a desired application pressure and target noz-
zle flow rate (during speed changes) by increasing or de-
creasing the on time of the nozzle-mounted solenoid. How-
ever, limited knowledge exists regarding the nozzle pres-
sure stability while turning nozzles on (energizing) and
shutting nozzles off (de-energizing) and the impact of noz- Figure 1. The VisiSize Imager measures particle diameter, shape, and
zle solenoid on time on spray coverage. Therefore, this velocity distribution for the cross-section of a spray fan pattern.

316 TRANSACTIONS OF THE ASABE


Table 1. Measured particle speed, percent volume, particle size, and
percent of total droplets for lateral distribution of spray droplets. 7%

Percent of Total Deposit


Speed Percent of Particle Percent 6%
Interval Deposit Size Range of Total 5%
-1
(m s ) Volume (m) Droplets 4%
0 to 3 27.9 <190 92.1
3%
>3 to 5 15.7 190 to 258 4.7
>5 to 7 14.4 258 to 336 1.8 2%
>7 to 9 23.3 336 to 500 1.1 1%
>9 to 11 18.6 500 to 925 0.3 0%
>11 to 13 0.2 >925 0.0 -100 -50 0 50 100
Lateral Deposit Spacing (cm)
Results showed that droplets smaller than 190 m com-
prised 92.1% of the particles and one-third of the deposit (a)
volume (table 1). Based on the ASABE standard for droplet
35%
size classification (ASABE, 2013), 99.7% of the particles

Percent of Total Deposit


30%
were found to be extremely fine to very coarse, and 81.3%
25%
of the deposit volume was comprised of droplets smaller
20%
than 500 m.
15%
In addition, the spray pattern of the selected nozzle type
10%
was quantified for lateral and longitudinal static deposits
using a spray pattern check tray at Teejet Spraying Systems 5%
in Glendale Heights, Illinois. The check tray was 5 m wide 0%
-30 -20 -10 0 10 20 30
and 2 m in depth with 5 cm partitions to capture spray depos-
its in the lateral and longitudinal directions (fig. 2). With the Longitudinal Deposit Spacing (cm)
partitions providing 5 cm ground resolution, the empirical
droplet distributions in the lateral and longitudinal directions (b)
(fig. 3) were characterized for use in development of the Figure 3. Characteristic (a) lateral and (b) longitudinal application
dynamic simulation software, as described later. distributions of a Wilger MR110-05 nozzle.

SETUP FOR MONITORING FLOW AND PRESSURE controller system (Viper 4, Raven Industries, Sioux Falls,
The setup for monitoring the system flow and pressure S.D.) used a flowmeter (15P, Raven Industries) and butter-
used an all-terrain vehicle (Mule 3010, Kawasaki Motors fly valve (063-0171-894, Raven Industries) to regulate the
Corp. USA, Lincoln, Neb.) equipped with a 6.6 m, three- product flow rate and pressure. A Capstan Pinpoint control-
section boom sprayer consisting of 13 nozzles (Wilger ler was used to set the desired system pressure, set the ac-
MR110-05 combo-jets) spaced at 50.8 cm intervals (fig. 4). tive nozzle configuration, and program the desired duty
The nozzles were numbered 1 through 13 starting from the cycle.
left boom section. The sprayer used a centrifugal pump
powered by 3.6 kW gasoline engine (GX160, Honda En- EXPERIMENTAL DESIGN
gines Group, Alpharetta, Ga.) to pressurize the system. The For each duty cycle and nozzle configuration test (de-

Figure 2. Teejet Industries spray pattern check tray with 100 slots measuring 5 cm wide with 2 m depth.

60(2): 315-325 317


Figure 4. Sprayer setup with Capstan Pinpoint control system and sensors (i.e., pressure transducers, flowmeter, and solenoid signal power).

scribed below), thin-film membrane pressure transducers assembly (fig. 5) did not impact the pressure or spray fan
(1502 B81 EZ 100 PSI G, PCB Piezotronics, Inc., Depew, pattern latency. The sprayer boom was set at a height of
N.Y.) were used to record pressure in the left, middle, and 53.3 cm, as recommended for PWM nozzles with a 110°
right boom sections, nozzle pressure of an even-numbered spray fan pattern (Capstan, 2013). The spray system was
nozzle (nozzle 2) and an odd-numbered nozzle (nozzle 3), programmed to apply 94 L ha-1 at duty cycles of 20%, 40%,
and system pressure (fig. 5). Preliminary tests were per- 60%, 80%, and 100% with 1, 3, 8, and 13 active nozzles at
formed to observe the spray response from nozzles with a constant target pressure of 345 kPa.
and without the pressure transducer adapter fittings using a A data acquisition system (Compact Rio, National In-
high-speed camera (model acA640-750uc, Basler, Inc., struments, Austin, Tex.) was used to record real-time pres-
Exton, Pa.). The high-speed camera was operated at sures, on/off nozzle solenoid signals, and flow rate. A pro-
750 frames per second. After both nozzles were signaled gram was develop in LabView (National Instruments, Aus-
on, the camera was used to determine the time necessary tin, Tex.) to count the rising and falling edge of the sole-
for liquid to exit the nozzles and form the spray fan pattern. noid pulses (for nozzles 1 and 2) during each test through
The results showed that the pressure transducer and fitting the respective solenoid valve to cross-check the applied

Figure 5. Capstan Pinpoint control system and pressure transducers installed to measure boom and nozzle pressures.

318 TRANSACTIONS OF THE ASABE


duty cycle. The data acquisition system sampled the digital vp = velocity of spray droplet (m s-1)
and analog sensor output at 1000 Hz and recorded the data t = deposit time (ms)
in a *.xlsx file for analysis.
Actual nozzle application uniformity was recorded using vs  0.02778  t  xd (2)
the spray pattern check tray for each test in order to com- where
pare the dynamic simulation results, described in the next xd = on-ground deposit distance from nozzle (cm)
section, to the static application. The check tray, with vs = sprayer speed (km h-1)
23 slots measuring 5 cm wide each, was used to catch up to t = deposit time (ms).
100 mL of spray. The number of pulses per test and the The deposit times were computed for lateral distance in-
amount deposited were used to determine the amount of crements of 5 cm to the left and right of the center of the
product deposited (L ha-1) per pulse cycle (100 ms). Each nozzle. These distance increments represented the slot loca-
test was run for 30 s, and the volume in each slot of the tions on the spray pattern check tray (fig. 2).
check tray was measured to calculate the average applica-
tion rate. Step 2: Product Volume and Area Covered
The flow characteristics were combined with the real-
DYNAMIC SIMULATION SOFTWARE time nozzle pressure to determine the real-time product
Dynamic simulation software was developed using Ex- flow (L min-1) from each nozzle. A cubic polynomial was
cel (Microsoft Corp., Redmond, Wash.) and Matlab (The used to characterize the flow (L min-1) versus nozzle pres-
MathWorks, Inc., Natick, Mass.). The software generated sure (kPa) in order to have the y-intercept pass through
on-ground spray coverage for each duty cycle (i.e., 20%, (0,0), as shown in figure 7, because no pressure would re-
40%, 60%, 80%, and 100%). The software computed pa- sult in no flow. The sprayer travel speed (km h-1) was cal-
rameters using steps 1 through 4, as described in the fol- culated based on the sprayer duty cycle per the manufactur-
lowing sections. er’s recommendations (fig. 8). The Capstan PWM control-
ler used a programmed 94 L ha-1 application rate and prede-
Step 1: Droplet Deposit Time and termined duty cycle to generate the expected ground speed
On-Ground Deposit Location of the sprayer (fig. 8).
For the dynamic simulations, it was assumed that the The sprayer travel speed (km h-1) was multiplied by the
spray particle speed remained constant during travel in the slot width on the spray pattern check tray (5 cm) to deter-
sprayer travel direction (x-direction), downward direction mine the on-ground coverage area per unit of time (ha min-1)
(z-direction), and lateral direction (y-direction). Two- (eq. 3):
dimensional particle kinetics were used to determine the
droplet deposit time and on-ground longitudinal
(x-direction) deposit location. Six droplet speed intervals 3.0
and the respective particle size ranges were used to com-
pute these two parameters (table 1). Each of the six speed 2.5
Flow Rate (L min-1)

intervals, the nozzle height, and the lateral and longitudinal 2.0
distributions (fig. 3) were used to compute the droplet de-
posit time (fig. 6 and eq. 1) and on-ground longitudinal 1.5
deposit distance (x-direction) in the simulation program 1.0
using equation 2: y = 2E-08x3 - 2E-05x2 + 0.0123x
R² = 0.9958
0.5

(h 2  y 2 ) 0.0
 1000  t (1) 0 100 200 300 400 500 600
vp
Pressure (kPa)

where
Figure 7. Flow versus pressure of Wilger MR110-05 nozzle (adapted
h = boom height (m) from Wilger, 2017).
y = on-ground distance from nozzle center (m)
25
800
Deposit Time Delay (ms)

h-1)

20
700
Travel Speed (km

600 0-3 m/s 15


500
3-5 m/s
400 10
300 5-7 m/s
200 5
y = 21.324x + 1.3679
100 7-9 m/s
0 0
-50 -40 -30 -20 -10 0 10 20 30 40 50 0% 20% 40% 60% 80% 100%
Lateral Spacing (cm) Duty Cycle (%)

Figure 6. Characteristic deposit latency of droplets at six speed inter-


vals, where 0 represents the center of the spraying nozzle. Figure 8. Speed versus duty cycle (adapted from Capstan, 2013).

60(2): 315-325 319


Table 2. Sprayer travel speed, distance traveled, and spatial coverage
area based on the respective duty cycle.
Duty Sprayer Sprayer Distance On-Ground Area
Cycle Travel Speed Traveled per Unit of Time
(%) (km h-1) (cm s-1) (ha min-1)
20 5.63 156.5 4.76E-04
40 9.90 274.8 8.40E-04
60 14.16 393.4 1.20E-03
80 18.42 511.8 1.56E-03
100 22.69 630.4 1.92E-03

a  v  8.47  105 (3)


where
a = on-ground area coverage (ha min-1) Figure 9. Simple representation of the combined longitudinal and
v = sprayer travel speed (km h-1). lateral distributions with respect to sprayer travel direction in the
dynamic simulation software.
The on-ground spatial area traveled per unit of time (ha
min-1) was input directly into the simulation software, as
smaller droplets, as shown empirically in table 1. Part 2 of
shown in table 2.
the simulation used the volume from each lateral spatial
Step 3: Application Rate and Spray Coverage increment in part 1 and redistributed the deposit into longi-
The real-time product flow rate (L min-1) computed tudinal spatial increments based on the longitudinal distri-
from the pressure response in step 2 was combined with the bution and particle speed (figs. 3b and 6).
area covered per unit of time (ha min-1) to calculate the
application rate (L ha-1), as specified in equation 4:
Step 4: Spray Distribution Map
8 3 5 2 The dynamic spray distribution results were then super-
2 p 2 p  0.0123p
R (4) imposed to represent the longitudinal and lateral nozzle
u overlaps. Based on the lateral nozzle spacing of 50.8 cm
where and the odd and even PWM solenoid offset of 50 ms, the
R = application rate (L ha-1) simulation program generated a dynamic on-ground appli-
p = real-time nozzle pressure (kPa) cation deposit for each duty cycle. The spray distribution
u = on-ground area per unit of time (ha min-1). from five consecutive nozzles was used to generate on-
For spray distribution, the longitudinal and lateral distri- ground coverage maps and conduct analysis of the as-
butions quantified using the spray pattern check tray applied application accuracy.
(figs. 3a and 3b) were combined to simulate the deposit
from a fully developed fan pattern (fig. 9). Based on the
lateral and longitudinal characteristic deposits shown in RESULTS AND DISCUSSION
figure 9, the combination of longitudinal (for- The on/off latency was consistently found to be 20 ms
ward/rearward) and lateral overlap percent flow is shown in before the nozzle reached the desired system pressure and
figure 10. A small percentage of the flow is deposited to- before the nozzle fully depressurized (fig. 12). In addition,
ward the outer perimeter of the fan pattern. once the nozzle solenoid was de-energized, the nozzle con-
The dynamic simulation was divided into two parts tinued spraying at the desired system pressure for 10 ms.
(fig. 11). Part 1 used the flow from the nozzle based on Overall, a fully developed spray fan pattern was produced
1 ms of on time using the duty cycle versus pressure data at the desired pressure for less time than the specified duty
collected at 1000 Hz. The computed flow was then distrib- cycle (table 3).
uted in lateral spatial increments using the lateral distribu- In practice, a duty cycle shorter than 33% is not recom-
tion (fig. 3a) and the computed on-ground deposit time mended for commercial PWM spray systems. As shown in
delay (fig. 6). As shown in figure 11, larger droplets have a figure 12, a duty cycle shorter than 50% reduced the
larger percent volume and travel faster when compared to odd/even nozzle signal overlap, resulting in operating times

Figure 10. Example of combined distributions of lateral and longitudinal flow characteristics at 345 kPa.

320 TRANSACTIONS OF THE ASABE


Figure 11. Simplified diagram of the two-part dynamic coverage combining the lateral (part 1) and longitudinal (part 2) flow distributions.

350 1
Pressure (kPa)

Solenoid Signal
280 Odd Solenoid
210 Even Solenoid
140 20% Odd Signal
70 20% Even Signal
0 0
0 20 40 60 80 100 120 140 160 180 200
Cycle Time (ms)
(a) 20% Duty Cycle

350 1

Solenoid Signal
Pressure (kPa)

280 Odd Solenoid


210 Even Solenoid
140 40% Odd Signal
70 40% Even Signal
0 0
0 20 40 60 80 100 120 140 160 180 200
Cycle Time (ms)
(b) 40% Duty Cycle

350 1
Solenoid Signal
Pressure (kPa)

280 Odd Solenoid


210 Even Solenoid
140
60% Odd Signal
70
60% Even Signal
0 0
0 20 40 60 80 100 120 140 160 180 200
Cycle Time (ms)
(c) 60% Duty Cycle

350 1
Solenoid Signal
Pressure (kPa)

280 Odd Solenoid


210 Even Solenoid
140 80% Odd Signal
70 80% Even Signal
0 0
0 20 40 60 80 100 120 140 160 180 200
Cycle Time (ms)
(d) 80% Duty Cycle
Figure 12. Pressure response following on/off nozzle power signals with duty cycles of (a) 20%, (b) 40%, (c) 60%, and (d) 80%.

60(2): 315-325 321


Table 3. Observed time with nozzle spray at pressure within 10% cause the system to react with significant differences, the
target for different duty cycles. characteristic flow from a single nozzle could be used to
Actual Time within Percent Time outside
simulate the on-ground coverage with the developed simu-
10% Target Pressure 10% Target Pressure
Duty Cycle (ms) (%) lation program. For practical applications, this observation
20% DC 14 30.0 supports the system’s ability to activate and deactivate mul-
40% DC 33 17.5 tiple nozzles and maintain the target pressure at all times.
60% DC 52 13.3
80% DC 74 7.5 SPRAY DISTRIBUTION AND COVERAGE MAPS
100% DC 82[a] - Figure 14 shows an example of empirical static spray
[a]
Measured when energized for the first time.
deposit using the spray pattern check tray. The spray depos-
it results showed that nozzle overlap resulted in peaks at
when the sprayer was not at the desired pressure. Spraying -50.8, 0.0, and +50.8 cm lateral spacing. The tests were
when the nozzles are not at the desired pressure could result conducted in a laboratory; therefore, there was no effect of
in application errors. When comparing the on time versus wind on droplet movement and deposition, which could
the duty cycle, inaccurate nozzle pressure occurred for a occur in field settings. All test results indicated that the
greater percentage of time with shorter duty cycles static application rate deposited on the pattern check tray
(table 3). Although the ground coverage area is proportion-
was within 5% of the target application rate of 94 L ha-1
ally smaller with shorter duty cycles, these errors can im-
(table 4). This illustrates the difference between static and
pact the spray coverage uniformity. Overall, the results
dynamic application accuracy.
indicated that application errors could occur when operat-
The dynamic theoretical on-ground spray coverage gen-
ing PWM spray systems with shorter duty cycles.
erated by the simulation software is shown in figure 15.
Monitoring the system (fig. 4) showed that the pressure
The simulated spray coverage for each duty cycle was cre-
remained within 5% error regardless of the number of ated at the same scale to represent the on-ground coverage
active nozzles configured for the test (fig. 13). This obser- from an overhead, or nadir, view. The simulation results
vation was attributed to the system’s ability to compensate showed that the product flow rate (L) per pulse was within
for changing product flow with varying active nozzle con-
the 5% error range except for the 20% duty cycle, which
figurations while maintaining the desired system pressure
is below the manufacturer’s recommended duty cycle.
of 345 kPa. Because the number of active nozzles did not

60
414
Target Pressure 1
+5% Error 1N1 60%
Nozzle

50
345
-5% Error Solenoid ON Signal
3N3 60%
Nozzles
Pressure (kPa)

40
276
8N8 60%
Nozzles
30
207
13N 60%
13 Nozzles
20
138
60% DC
On Signal
1069 Signal

00 0
0 20 40 60 80 100
Cycle Time (ms)

Figure 13. On/off latency with different numbers of active nozzles. The black dotted line represents the target pressure.

120
100
80
L ha-1

60
40
20
0
-60 -40 -20 0 20 40 60
Laterial Spacing (cm)

Figure 14. Static spray droplet distribution on the spray pattern check tray at 345 kPa with an 80% duty cycle.

322 TRANSACTIONS OF THE ASABE


Table 4. Empirical application rates (L ha-1) for each duty cycle based coverage maps indicate the need for future studies with
on droplet distribution on the spray pattern check tray. PWM systems to improve spray coverage.
Duty Cycle
To quantify the simulated results shown in figure 15, the
20% 40% 60% 80% 100%
results were graphed to show the percent of coverage area
Average 91.8 93.8 92.9 98.6 94.4
(C = 95%) 5.0 3.7 3.1 3.2 2.3 versus application rate (fig. 16). As shown in figure 16, the
Error -2.34% -0.21% -1.17% 4.89% 0.43% higher duty cycles (i.e., greater than 50%) resulted in a
higher percentage of area receiving the target application
However, the simulated coverage maps showed areas that rate. For example, the 80% duty cycle resulted in a more
received over- and under-application (fig. 15). Figure 16 even distribution of product applied surrounding the target
shows the percentage of areas that received different appli- rate of 94 L ha-1.
cation rates. The results indicated that the on-ground appli- The resulting application accuracy was segmented into
cation rates varied from 77 to 112 L ha-1 for simulated three categories (<5% below target, 5% of target, and
spray applications at different duty cycles. For duty cycles >5% above target rate), as shown in figure 17a, to visualize
above 20%, the average static coverage of each duty cycle the variability in application rate. For the five simulated
was within 5% (89.3 to 98.7 L ha-1) of the target product duty cycles, anything less than a 100% duty cycle resulted
rate of 94 L ha-1 (fig. 15); however, the simulated spray in significant areas receiving an application rate beyond
coverage was not uniform at all times. In practice, con- 5% of the target rate. Additionally, significant areas had
sistent over- or under-applications in certain areas can po- more under-application (35% to 79%) than over-application
tentially lead to ineffective use of pesticides as well as (1% to 17%). These higher instances of under-application
growth inhibitors and promoters. Overall, the simulated are similar to the results of earlier studies conducted on
flow-based systems (Sharda et al., 2012, 2011). In addition,

2
*Area = 1.92 m

18 37 56 74 94 112 131 2
*Area = 1.56 m

L ha-1 2
*Area = 1.20 m
Direction of Travel

190.5 cm
152.4 cm
2
*Area = 0.84 m
116.8 cm

2
*Area = 0.48 m
81.3 cm
45.7 cm

101 cm

Average Coverage Average Coverage Average Coverage Average Coverage Average Coverage
84.00 L ha-1 89.80 L ha-1 90.55 L ha-1 91.48 L ha-1 94.29 L ha-1
20% Duty Cycle 40% Duty Cycle 60% Duty Cycle 80% Duty Cycle 100% Duty Cycle
Figure 15. Simulated spray coverage from three complete pulses of five consecutive nozzles with 20%, 40%, 60%, 80%, and 100% duty cycles.

20% DC 40% DC 60% DC 80% DC 100% DC


Percent of Applicatin Rate

40%
35%
30%
25%
20%
15%
10%
5%
0%
74.8 77.2 79.5 81.8 84.2 86.5 88.9 91.2 93.5 95.9 98.2 100.6 102.9 105.2 107.6 109.9 112.2
Application Rate (L ha-1)

Figure 16. Percent of coverage area versus application rate for simulations with 20%, 40%, 60%, 80%, and 100% duty cycles (DC).

60(2): 315-325 323


5% Below Target ±5% Desired Target 5% Above Target

100% 1%
90% 17% 17% 14% 13%
20%

As Applied Application
80%
70% 25% 31%
60% 50%
50%
40% 84%
79%
30% 57% 51%
20% 35%
10%
0%
20% 40% 60% 80% 100%
Duty Cycle

(a)

10% Below Target ±10% Desired Target 10% Above Target

100% 0% 0%
8% 8%
90%
As Applied Application

80% 38%
70%
60% 67%
50% 77% 94% 100%
40%
30% 61%
20%
10% 24%
14%
0% 6%
20% 40% 60% 80% 100%
Duty Cycle

(b)
Figure 17. Deposit results for the five duty cycles represented with an acceptable errors (a) 5% and (b) 10%. The as-applied simulation
results show on-ground application rates for each of the duty cycles.

a 100% duty cycle produced higher application accuracy, CONCLUSION


where the as-applied application rate matched the target A full-system monitoring setup, high-speed imagery,
application rate more closely, because the nozzles were and particle analysis were combined to create a test proto-
operating under “conventional” spraying conditions. col to characterize nozzle flow and deposition for dynamic
The application accuracy is shown in figure 17b by simulation of on-ground spray coverage. Results showed
segmenting the accuracy with a 10% error. By setting an that the commercial PWM spraying system was able to
acceptable range of 10%, the simulated spray coverage maintain the desired boom pressure regardless of the num-
exhibited nearly uniform coverage for the 80% and 100% ber of active nozzles. The on/off latency in pressure devel-
duty cycles and greater accuracy in the as-applied applica- opment was consistently found to be 20 ms before the noz-
tion for the other duty cycles by reducing under- and over- zle reached the desired system pressure and before the noz-
application areas. However, under- and over-application zle fully de-pressurized. In addition, when the nozzle sole-
still occurred for duty cycles less than 80% (fig. 17b). noid was de-energized, the nozzle continued spraying at the
Overall, the results indicated that further research should be desired system pressure for 10 ms.
conducted in simulated and actual field conditions to un- Although the system maintained the desired system
derstand the efficacy of PWM systems in applying chemi- pressure, the on time specified by the duty cycle was not
cals. The current study was conducted with the manufac- achieved with the PWM system, thereby contributing to
turer’s recommended boom height and with one nozzle areas with spray coverage that was outside the target range.
type. Further studies should consider different boom The static spray pattern collection showed that the PWM
heights and nozzle types to develop best management prac- system provided adequate flow within 5% of the target
tices for end users. range with every pulse, but the as-applied simulated spray

324 TRANSACTIONS OF THE ASABE


coverage did not show uniform application. http://dx.doi.org/10.13031/2013.26525
The high-speed image analysis results showed that drop- Luck, J. D., Pitla, S. K., Shearer, S. A., Mueller, T. G., Dillon, C.
lets smaller than 190 m comprised 92.1% of the particles R., Fulton, J. P., & Higgins, S. F. (2010). Potential for pesticide
and nutrient savings via map-based automatic boom section
and one-third of the deposit volume (table 1) with a deposit
control of spray nozzles. Comput. Electron. Agric., 70(1), 19-26.
delay time of 600 to 700 ms (fig. 6). Simulation analysis http://dx.doi.org/10.1016/j.compag.2009.08.003
revealed that higher duty cycles resulted in increased cov- Luck, J. D., Sharda, A., Pitla, S. K., Fulton, J. P., & Shearer, S. A.
erage efficiency compared to lower duty cycles. Sprayer (2011b). A case study concerning the effects of controller
PWM systems with duty cycles above 60% would provide response and turning movements on application rate uniformity
greater spray coverage uniformity. In addition, PWM duty with a self-propelled sprayer. Trans. ASABE, 54(2), 423-431.
cycles above 60% would provide more overlap between http://dx.doi.org/10.13031/2013.36445
odd and even nozzles and increase the application efficacy. Luck, J. D., Zandonadi, R. S., & Shearer, S. A. (2011a). A case
Dynamic spray simulations showed that the target applica- study to evaluate field shape factors for estimating overlap errors
with manual and automatic section control. Trans. ASABE,
tion error may vary beyond 5%. At the same time, the 54(4), 1237-1243. http://dx.doi.org/10.13031/2013.39022
product flow rate per pulse was within the acceptable lim- Porter, W. M., Rascon, J., Shi, Y., Taylor, R. K., & Weckler, P.
its. As such, while the PWM system provided the amount (2013). Laboratory evaluation of a turn compensation control
of product desired per pulse, the simulations suggested that system for a ground sprayer. Appl. Eng. Agric., 29(5), 655-662.
the on-ground coverage could result in areas with under- or http://dx.doi.org/10.13031/aea.29.10075
over-application. Further tests should be conducted to vali- Sama, M. P., Evans, J. T., Turner, A. P., & Dasika, S. S. (2016). As-
date the simulation results of this study with actual dynam- applied estimation of volumetric flow rate from a single sprayer
ic spray tests in field conditions. nozzle series using water sensitive spray cards. Trans. ASABE,
59(3), 861-869. http://dx.doi.org/10.13031/trans.59.11538
Sharda, A., Fulton, J. P., McDonald, T. P., & Brodbeck, C. J.
ACKNOWLEDGEMENTS (2011). Real-time nozzle flow uniformity when using automatic
Special thanks are extended to the Department of Bio- section control on agricultural sprayers. Comput. Electron.
logical and Agricultural Engineering at Kansas State Uni- Agric., 79(2), 169-179.
versity for the testing space, supportive funding, data ac- http://dx.doi.org/10.1016/j.compag.2011.09.006
quisition hardware and software, guidance, and support to Sharda, A., Fulton, J. P., McDonald, T. P., Zech, W. C., Darr, M. J.,
conduct this research. Similarly, Capstan’s support during & Brodbeck, C. J. (2010). Real-time pressure and flow dynamics
this research was greatly appreciated for correctly configur- due to boom section and individual nozzle control on
ing the sprayer for our respective tests. agricultural sprayers. Trans. ASABE, 53(5), 1363-1371.
http://dx.doi.org/10.13031/2013.34891
Sharda, A., Luck, J. D., Fulton, J. P., McDonald, T. P., & Shearer,
S. A. (2012). Field application uniformity and accuracy of two
REFERENCES rate control systems with automatic section capabilities on
ASABE. (2013). S572.1: Spray nozzle classification by droplet agricultural sprayers. Prec. Agric., 14(3), 307-322.
spectra. St. Joseph, MI: ASABE. http://dx.doi.org/10.1007/s11119-012-9296-z
Capstan. (2013). Pinpoint Synchro product manual. Topeka, KS: Teejet. (2014). Teejet Technologies, catalog 51A-M. Wheaton, IL:
Capstan Ag Systems. Spraying Systems Co.
Grisso, R. D., Dickey, E. C., & Schulze, L. D. (1989). The cost of Wilger. (2017). Publication 10216-US. Lexington, TN: Wilger, Inc.
misapplication of herbicides. Appl. Eng. Agric., 5(3), 344-347.

60(2): 315-325 325

You might also like