Anda di halaman 1dari 13

J. agric. Engng Res.

(1989) 42, 135-147

A Simulation Model of the Spray Drift from Hydraulic Nozzles


P. C. H. MILLER*; D. J. HADFIELD?
A model to predict the trajectories of droplets from agricultural flat fan nozzles is
described. Droplet motion is considered in two phases: close to the nozzle where the
trajectory is dominated by the conditions associated with droplet formation, particularly the
initial velocity and entrained air conditions; and a second phase where the effects of
approach. Results from the
atmospheric turbulence are predicted using a random-walk
model are compared with laboratory measurements of droplet size/velocity profiles beneath a
nozzle and with field measurements of the downwind drift from boom mounted nozzles
operated conventionally. Predicted total downwind drift deposits agreed well with measured
values but the form of the vertical deposit distribution was less well predicted.

1. Introduction
Previous
work concerned with predicting spray droplet movement has used both
dispersion and random-walk type models. Dispersion models have been used to study the
transport of sprays both towards and within crop canopies and have been applied
particularly to aerial spraying where droplet movements through the air can be relatively
large. Bathe and Sayer12 and Cramer (Dumbauld et a1.)3 derived expressions for the
deposit distribution downwind of a line spray source and for the airborne flux just above
the crop canopy based on diffusion theory, and demonstrated some agreement with field
data. Schaefer and Allsop* discuss the application of gradient diffusion theory to spray
transport and demonstrate an improved prediction of deposit when compared with the
linearly expanding Gaussian plume used by Bathe and Sayer and by Cramer.
A number of authors have used Markov type simulation models to predict the
trajectories of droplets in turbulent air flows including Thompson and Ley, Picot et al. ,6
Wilson et al. ,7 and Legg and Raupauch.* Such models have been shown to provide an
acceptable description of spray deposit distributions downwind from a defined source
given assumed release conditions. These models have been used in conjunction with both
aerial and ground crop spraying systems and have examined the effects of operating
parameters on downwind spray deposits. Both diffusion and random-walk type models
have been developed that account for droplet evaporation
and sedimentation
by
gravitational forces. However, droplet release conditions have been assumed and not
related to sprayer characteristics and relatively little work has attempted to relate the
conditions of droplet generation to their subsequent transport particularly in the region
close to agricultural flat fan nozzles.
Such an analysis will provide a basis for the design of spraying systems to minimise drift
and is the subject of this paper.
Droplet trajectories are described in two distinct phases: close to the nozzle where
considerations of air drag and entrained air predominate, and further downwind where a
random-walk approach is used.

* Chemical Applications Group, AFRC Engineering, Silsoe, Bedford MK45 4HS, UK


7 Formerly of above address. Current address: Flat 3, 56 Chaucer Road, Bedford MK40 2AP, UK
Received 9

June

1988; accepted in revised form 6 November 1988.

Paper presented at Ag Eng 88, Paris, France, 2-6 March 1988.

135

136

A MODEL

OF

SPRAY

DRIFT

Notation
d
0,
h
K1
L
I,
r
s
T,
u,
u
24,
u,
v,
w
z

droplet diameter, urn


distance downwind of nth sampling line, m
distance below nozzle, m
entrained air parameter,
dimensionless
Monin-Obukov
length, m
coherent length of liquid spray
sheet, m
see Eqn (3)
nozzle spacing on boom, m
Stokesian response time, s
sheet velocity for fan nozzle, m/s
horizontal droplet velocity, m/s
friction velocity, m/s
entrained air velocity, m/s
droplet settling velocity, m/s
vertical droplet velocity, m/s
height above ground, m

z, crop height, m
z, boom height, m
(Y factor indicating loss of velocity
correlation in successive time steps
y see Eqn (3)
6 constant in entrained air equation
rr_ Lagrangian time scale, s
At time step, s
Ap vapour pressure difference, Pa
& random variables from Gaussian
rl 1 distribution with mean zero and
standard deviation of 1-O
p viscosity of air
p density of droplet, kg/m3
velocity fluctua0 rms horizontal
tions, m/s
%I rms vertical velocity fluctuations,
m/s

of the model

2. Structure

The model simulates a simplified spraying situation as shown in Fig. 1 in which the
drift from a set of boom mounted nozzles operating above a cereal crop canopy is
measured using ve&al strings mounted at various distances downwind. In structuring the
model the following simplifying assumptions were made:

spray

(1) Trajectories

Wind

were considered in two dimensions only and effects due to the forward

direction
Sampling
lines

c-----D,
.

Fig. 1. Simulated spraying conditions

P. C. H. MILLER;

137

D. J. HADFIELD

movement of the sprayer and the turbulent wakes created by this forward
movement were neglected.
(2) The crop surface behaved as a perfect collector of spray with no movement or
reflection of droplets from within the canopy. Ground spray deposits were
calculated on this basis.
(3) Spray droplets from adjacent nozzles on a boom acted independently.
2.1. The random-walk model
The random-walk model used in the simulation was based on that developed by
Thompson and Leg in which the velocity of a droplet at any time step was related to the
velocity in the previous time step but with an added random component due to
turbulence. The droplet velocity in the vertical direction, w, in the i + 1 th time step is
given by:5
wi+l

(r(Wi

+ Vsi) +

qj+lUw(l-

0!2)t - us,+,

(1)

where Q = exp(-At/r,_)
and 7 is a random variable. Eqn (1) is only valid if the time step
At is small compared with the Lagrangian time scale, rL. The effect of droplet settling
velocity, u,, is included as a direct term in the equation, and this was considered by
Thompson and Ley to be satisfactory for water based droplets with diameters up to
450 urn based on data by Smith. Settling velocities, IJ,~,are therefore calculated using the
relationships given by Thompson and Ley as
v,, = 4.47 x 10-3d - 0.191
for d>lOOum
and

(2)

v,, = 3.2 x 10+d2 - 6.4 x 10+d3


for d<lOOum
The horizontal droplet velocity, U, in the direction of the wind is given5 by:
ui+l

(yui

U~*,+l,(l- (u) + yU(l - 2)f

where
y = 05rqi+l + (1 - 125r2)&i+1
and

(3)
(4)

2
r=-

u*
(J G!

The trajectories of droplets over a series of time steps of At were calculated based on
Eqns (l)-(4) once an effective release position for each droplet had been reached (see
Section 2.2).
2.1.1. Droplet evaporation
The change in droplet diameter due to evaporation was based on relationships given in
Thompson and Ley assuming that mass transfer occurred as if the droplets were of
water, down to a minimum size of 1% of the original droplet volume. The change in
droplet diameter was determined from
Ad = - (0*358Ap/d)(2 + 0*124(v, - d)i) At
The effects of droplet density or evaporation
not considered in the model.

(5)
of pesticide materials in the spray were

138

A MODEL

2.2.

OF SPRAY

DRIFT

Trajectory calculation close to the nozzle

Trajectories were determined over a series of time steps using an integration routine
described by Marchant with drag coefficients equal to those for solid spheres. Marchant
indicated that errors due to this assumption would be less than 10% for the droplet
conditions simulated.
The effect of turbulence on initial droplet trajectories was simulated by including an
additional vertical air velocity component term which was sampled randomly from a
Gaussian distribution with a mean of zero and a standard deviation of a, and which was
assumed to persist for a period equal to the Langrangian timescale, r,_.
The response time of a droplet in a moving air-stream is characterised by the Stokesian
response time, T, where

Initial runs with the model indicated the droplets below a spray nozzle had reached
their settling velocities after a period of approximately 4T, and so this value was used as
the basis for changing from the ballistic trajectory to random-walk phases of the model.
2.2.1. Spray characteristics
Spray droplets were assumed to form at a distance equal to the sheet coherent length
below the nozzle and have an initial velocity the same as the liquid sheet. Data relating to
coherent lengths and sheet velocities was initially obtained from measurements made
using high-speed photography.
Trajectory angles were sampled from a Gaussian distribution with a mean of zero and a
standard deviation of O-4 times the nozzle angle. This value was determined experimentally by comparing calculated volume distributions below a 110 flat fan nozzle with those
measured in patternator experiments.
The effect of the droplet size distribution was accounted for by simulating the
trajectories of between 500 and 5000 droplets in each of a number of size categories
(depending on the required accuracy and limitations of computing time) up to a maximum
of 25 size categories. Size categories in increments of 20 urn have been found convenient
in applying the simulation to most agricultural flat fan nozzles. Volume distributions were
then determined by relating the proportion of droplets deposited or remaining airborne in
a defined area, to the total volume of spray liquid output while travelling through the
sample volume and the measured droplet size/volume distribution for the spray produced
by the nozzle.

2.2.2. Entrained air velocities


Studies of air entrainment in liquid sprays have shown that the velocity of entrained
along the axis of a fan jet nozzle, v,, can be described by the equation:12

where u, is the sheet velocity and h the distance below the nozzle. 8
for sprays into air takes a value* of 0.4, while K1 is a constant defined
spray fan at right angles to the spray sheet at a given distance from
determined from photographic measurements of a number of 80 and

air

is a constant which
by the width of the
the nozzle. K1 was
110 nozzles, and a

P. C. H. MILLER;

139

D. J. HADFIELD

Start

Input nozzle
conditions

Initiate droplet
flight

-8

Fig. 2. Model flow chart

140

A MODEL

mean value of 0.14 was used as input to the simulation.


assumed to be constant across the spray sheet.

OF SPRAY

The entrained

DRIFT

air velocity was

2.2.3. Structure of the simulation model


A flow chart for the model of drift from a single nozzle is shown in Fig. 2. The drift
from boom mounted nozzles was simulated by calculating the drift from a single nozzle
and overlapping that from successive nozzles off-set by the nozzle spacing distance s.
Inputs to the model in addition to those defining the spray nozzle included a description
of the sampling system and parameters defining the meteorological conditions. Atmospheric stability was defined in terms of the Monin-Obukov
length (L), and related to a,,
a, and tr by relationships derived in Thompson and Ley.5
Simulation outputs gave the vertical distribution of airborne spray in 10 cm increments
at each sampling line and the ground deposits between sampling lines.
The predicted trajectories of ten 100~urn droplets released in the spray from a flat fan
nozzle spraying at the rate of 0.6 l/min at a pressure of 3 bar are shown in Fig. 3.

1 Nozzle
; . ::

Crop

Fig. 3. Simulated trajectories of l&I pm droplets from a 110 nozzle operating 0.5m above a cereal
from random walk model
crop 50mm tall. . . . . . . -, from ballistic trajectory model; -,

3. Experimental
3.1. Measurements

verification

of the model

of downwind drift from a moving boom

Measurements of the drift from three boom mounted nozzles were made using the
techniques reported by Sharp. l3 The nozzles (Lurmark 110015 in Kemetal) were mounted

P. C. H. MILLER;

141

D. .I. HADFIELD

on a small boom and off-set from the axis of the tractor to minimise air disturbance and
wake effects. The nozzles sprayed a 1% solution of a fluorescent tracer dye at a pressure
of 3.0 bar in a total of 20 passes down a spray track aligned at right angles to the mean
wind direction so as to accumulate measurable deposits. A spraying speed of 8 km/h was
used. Spray drift was captured using 2.5 mm diameter plastic tubing suspended in a
3 x 6 m framework. Meteorological conditions at the time of spraying were recorded
using a 10 m mast with four vane anemometers, mean air temperature, relative humidity
and temperature difference sensors. The Monin-Obukov
length was derived from the
Richardson number using the method described by Thompson and Ley5 for conditions
above a cereal crop with Richardson number determined directly from weather mast
measurements. The droplet size spectrum from the nozzles was determined from data
collected from a Particle Measuring Systems size analyser and a laboratory x-y sampling
arrangement. l4
To simulate the drift from the boom arrangement of nozzles, the velocity of the liquid
sheet (and hence the initial velocity of the droplets) was estimated from data obtained
from high speed photography of similar nozzles operating at the same pressure and with
the same spray liquid.
Fig. 4 shows measured and simulated total line deposits at distances up to 6 m
downwind for two different weather conditions. Simulated values were calculated using
500 droplets in 20 size categories with a 20 urn increment between categories. The effects

360 320280 3
f

240-

5
,

C
I=
02

200160120EO-

,
----.
I

o
0

, --*-*-.
I\!_!_,

- --.-_.

4
DMance

!_!_

downwmd,

6
rn

Fig. 4. Measured and predicted downwind drift profiles from a moving boom for two atmospheric
measured in mean wind speed of 6*6m/s at a height of 1Om; - - - - -, predicted
conditions. -,
in mean wind speed of 6*6m/s at a height of 10m; - I - i - 1-, measured in mean wind speed of
1_9m/s at a height of 10m; - * - -, predicted in mean wind speed of 1_9m/s at a height of 10m

142

A MODEL

of entrained air were not included in this initial simulation


significantly over-estimated drift as expected.
3.2. Droplet size/velocity

OF SPRAY

DRIFT

and model predictions

profiles below a nozzle

Measurements of droplet size and velocity were made 0.5 m below the central axis of a
flat fan nozzle (Lurmark 110015 in Kemetal) spraying water plus 0.1% Agral (ICI plc)
using a Particle Measuring Systems size analyser,14 and the results are plotted on Fig. 5.
Data are for six replicated scans of the spray produced by the nozzle. Also plotted on Fig.
5 are the droplet size/velocity profiles calculated using the ballistic trajectory routine in
the model and on the following assumptions:
(1) a constant entrained air velocity between nozzle and sampling point equal to the
mean measured velocity of droplets in the 40-80 pm size range, and
(2) entrained air velocities defined by Eqn (7), and with values for the constants as
given in Section 2.2.2 (i.e d2/2K1 = O-57).
Velocities of the larger droplet size are more variable because of the relatively small
number of droplets of this size in the spray sample.
The results show that the velocities predicted using the trajectory routine in the model
were some 2-5 m/s above those measured and indicate that the initial (or liquid sheet)
velocity used in the simulation at 17.0 m/s was too high. It was also expected that the
theoretical plot in Fig. 5, based on a constant entrained air velocity between nozzle and
measuring point, would under-predict droplet velocities, and the fact that this was not the
case except above 450 pm also indicated that the release velocity was less than that
assumed.

0
Droplet

size, pm

Fig. 5. Measured and predicted droplet velocities below 110 nozzle. *, measured values; -,
predicted using trajectory routine in model with entrained air velocity determined from Eqn (7,);
----, predicted assuming a constant entrained air velocity

P. c.

H. MILLER;

143

D. J. HADFIELD

14-

12-

I
100

I
300

I
200
Droplet

sue,

I
400

pm

Fig. 6. Measured and predicted droplet velocities below a 110 nozzle with different entrained air
conditions. -,
62/.2KI in Eqn (7) = O-57; *, measured values; -. - . -, Sz/2K, in Eqn
(7) = O-95; - - - - - , 6 /2K, in Eqn (7) = 1.25

350-

300Ti
g 250b
$
D
E
L
z
t

200150IOO50-

0:

0.5

I.5

2.5
Distance

3.5
downwmd,

4.5

5.5

c 5

Fig. 7. Measured and predicted downwind drift profiles. *, measured; -,


predicted as Fig. 4;
---, predicted with entrained air, sheet velocity = 17.0 S2/2K, = 0.57; - . - . - , predicted with
entrained air, sheet velocity = 15.0 6 2/2K, = 0.95

144

A MODEL

OF

SPRAY

DRIFT

200
180-I

180
160

60.
40.

i$,,,
I

>

20.
0:

0.
2'0

4b

6b

Line deposit,

8b

Id0

I;

20

/II

40

6b

80

Id0

Line deposit, ~1

Fig. 8. Measured and predicted airborne spray profiles at 1 m (left) and 6m (right) downwind of a

predicted with no entrained air sheet velocity = 17aOm/s,


boom sprayer. *, measured; -,
- - - - , predicted with entrained air sheet velocity = 17.0 6 /2K, = O-57; - . - . -, predicted with
entrained air sheet velocity = 15.0 6 /2K, = O-95

Fig. 6 shows the same experimental data but with velocity profiles predicted assuming a
sheet velocity of 15.0 m/s and values of 6/2K1, in Eqn (7) of O-57, O-95 and l-25. Using
a value of O-57 gave predicted droplet velocities up to 3-5 m/s above measured values, and
some variation in the values of 6/2K1 was considered justified because of the practical
difficulties in measuring K, from photographic records. Values of O-95 and l-25 gave an
improved fit to the measured data although the form of the measured relationship was
still not well predicted.

3.3. Revised simulation

of measured

results

The simulation of the field drift measurements


was re-run with an initial droplet
velocity of 15-O m/s. Fig. 7 shows the effects of using revised velocity and assumed
entrained air conditions on the predicted downwind drift with the equivalent vertical
airborne spray profiles at distances of 1 and 6 m downwind shown in Fig. 8. The inclusion
of entrained air reduced drift volumes as expected with the best agreement between
measured and predicted drift obtained with an entrained air parameter of O-085
corresponding to a value of d2/2K1 in Eqn (7) of O-95.

4. Discussion of results
4.1. Entrained

air effects

The results indicate the importance of the correct definition of sheet velocity and
entrained air conditions close to the nozzle. The use of entrained air parameters giving a

P. C. H. MILLER;

145

D. J. HADFIELD

value of 6/2K1 close to unity gave an improved agreement with both measured
downwind drift and size/velocity profiles below the nozzle. Original values of the
entrained air parameters were derived from photographic measurements of the spray
section and were probably inaccurate due to the difficulty in identifying spray boundaries
on such photographs. Further work is needed to define the effects of entrained air
conditions for a range of nozzle types and sixes and in different parts of the spray
structure. It can be seen from the results in Figs 5 and 6 that with an entrained air
parameter of 0.085 (i.e. S2/2K1 = O-95) the model underestimates the vertical velocities of
droplets below 150 pm in diameter and yet it can be seen from Figs 7 and 8 that
downwind drift values are not overestimated.
The droplet size/velocity profiles were
measured in the laboratory at a relatively low horizontal scanning speed of O-05 m/s, and
this velocity difference may affect the trajectories of drifting droplets. Further work needs
to examine the interaction between the spray sheet and surrounding air movements and
to improve the definition of entrained air effects.
4.2. Change from trajectory to random walk routine
This was examined by examining the airborne drift volumes at distances up to 6 m from
the nozzle when simulating drift with different integer increments of the Stokesian
response time, T, in the trajectory routine of the model, and the results are shown in Fig.
9. With values of less than 4, drift values were lower indicating that droplets entered the
random walk routine of the model with downward velocities greater than the settling
velocity.
4.3. General discussion
The agreement between measured and predicted drift profiles from 3 boom mounted
nozzles was relatively good particularly in view of some of the assumptions made relating
to nozzle movement. With entrained air, predicted downwind profiles, shown in Fig. 7,

4oo

,/._..__...........
.....................................................
....
t

01

3
Time

in trqectory

IO

routine, nT,

Fig. 9. Effect of increasing time in trajectov routine on the total airborne drift up to 6m from a
boom with three 110 nozzles. -,
drift at 6m; - . - * -, drift at 3 m; - - - - , drift at 2m;
. . . . . 9 drift at 1 m

146

A MODEL

OF SPRAY

DRIFT

decayed more rapidly than those measured, and this is probably related to the method
used for incorporating a settling velocity component in the random-walk model.
Further work has now examined this feature of such models, e.g. Walklate,
and
incorporating an improved description of heavy particle trajectories downwind should
improve model predictions.
Work is being conducted to compare the model predictions with measurements of the
drift from single static and boom mounted nozzles to eliminate the effects of movement
and any nozzle interactions.
The predicted drift profiles differed from those measured directly above the
crop/ground surface indicating that the assumptions made regarding droplet retention and
reflection were too simplified.
All the calculations assumed that droplets behaved as spherical particles. This
assumption was probably valid for droplets up to 350 urn in diameter, but it can be seen
from the results in Figs 5 and 6 that the velocities of larger droplets are underestimated
because of the effects of shape distortion.
5. Conclusions and proposals for future work
The model provides a useful description of the spray drift from agricultural hydraulic
nozzles, with reasonable agreement between measured and predicted downwind drift
profiles.
Further work is required to extend the range of conditions for which the model has
been validated. Work is also required to improve the description of entrained air effects,
to include the effects of forward speed including the generation of turbulent wakes from
the boom and spray vehicle, and to improve the calculation of heavy particle
trajectories.
Acknowledgements
Thanks are due to Mr A. P. Chick and Mr J. E. Pottage for assistance with the field experiments
and to Mr C. R. Tuck for assistance with the laboratory measurements of droplet size/velocity
profiles.

References
Beche,
*
a
4
5

D. H.; Sayer, W. J. D. Transport of aerial spray, I. A model of aerial dispersion.


Agricultural Meteorology 1975, 15: 257-271
Bathe, D. H.; Sayer, W. J. D. Transport of aerial spray. II. Transport within the crop canopy.
Agricultural Meteorology 1975, 15: 371-377
Dumbauld, R. K.; Rafferty, J. E.; Bjorkland, J. R. Prediction of spray behaviour above and
within a forest canopy. USDA Forest Service Report, 1977, TR-77-308-01
Schaefer, G. W.; Allsop, K. Spray droplet behaviour above and within a crop. Proceedings 10th
International Congress of Plant Protection 1983, 3: 1057-1065
Thompson, N. J.; Ley, A. J. Estimating spray drift using a random-walk model of evaporating
drops. Journal of Agricultural Engineering Research 1983, 28: 418-435
Picot, J. J. C.; Kristmanson,
D. D.; Basak-Brown,
N. Canopy deposit and off-target drift in
forestry aerial spraying. The effects of operational parameters. Transactions of the ASAE
1986, 29: 80
Wilson, J. D.; llmtell,
G. W.; Kidd, G. E. Numerical simulation of particle trajectories in
inhomogeneous turbulence. IV: Comparison of predictions with experimental data for the
atmospheric surface layer. Boundary Layer Meteorology 1981, 21: 443-463.
Legg, B. J.; Ranpach, M. R. Markov chain simulation of particle dispersion in inhomogeneous

P. C. H. MILLER;

g
lo

l2

D. J. HADFIELD

147

flows. The mean drift velocity induced by a gradient in Eulerian velocity variance. Boundary
Layer Meteorology 1982,24: 3-13
Smith, F. B. The turbulent spread of a falling cluster. In: Advances in Geophysics 6,
Atmospheric Diffusion and Air Pollution. London: Academic Press 1959, pp. 193-210
Mar&ant, 1. A. Calculation of spray droplet trajectory in a moving airstream. Research Note.
Journal of Agricultural Engineering Research 1977, 22: 93-96
Lake, J. R. Sprays, physical properties and deposition on cereals and weeks. PhD Thesis,
University of London, 1982
Briffa, F. E. J.; Dombrowski,
N. Entrainment
of air into a liquid spray. American Institute of
Chemical Engineers Journal 1966, l2(4): 708-717
Sharp, R. B. Comparison of drift from charged and uncharged hydraulic nozzles. British Crop
Protection Council, Proceedings of Conference on Pests and Diseases, 1984, pp. 1027-1031
Lake, J. R.; Dix, A. J. Measurements of droplet size with a PMS optical array probe using an
x-y nozzle transporter. Crop Protection 1985, 4(4): 464-472
Walklate, P. J. A Markov-chain particle dispersion model based on airflow data: Extension to
large water droplets. Boundary Layer Meteorology 1986, 37: 313-318

Anda mungkin juga menyukai