United States
Environmental Protection
Agency
Air and Energy
Engineering Research Laboratory
Research Triangle Park, NC 27711
&
Research and Development
EPA/600/S7-88/006 Nov. 1988
x°/EPA Project Summary
The Role of Rogue Droplet
Combustion in Hazardous
Waste Incineration
R.K. Srivastava, J.V. Ryan, and J.V. Roy
In the incineration of liquid haz-
ardous wastes, atomization quality
may limit destruction efficiency.
Large, nonmean droplets in a fuel
spray can pass through the flame
zone prior to complete evaporation
and may subsequently fail to burn
completely due to insufficient tem-
perature and/or flame radicals. A
study was conducted to develop a
predictive understanding of individ-
ual droplet trajectories in turbulent
diffusion flames.
The influence of droplet spacing
on the drag coefficient of individual
drops injected into a quiescent
environment has been determined
through measurements of trajec-
tories of single, monodlsperse, non-
evaporating droplet streams. Droplet
size, velocity, and spacing were
varied, yielding initial Reynolds
numbers (Re) ranging from 90 to 290
and initial droplet spacing to
diameter ratios (L/D) ranging from
1.7 to 1700. Data from 10 trajectories
were correlated using asymptotic
forms for CD oft
{CD (Re, L/D)}~n
= {CD (Re, L/D)}- n + {(CD)00 (Re)}~ n
where
(CD)°=CROD(Re) + aReb{L/D-l}
and is the asymptotic form for CD as
L/D->1, while (CD)" is that for CD as
L/D-><*>. A trajectory model contain-
ing the local drag coefficient was fit
to the experimental data by a
nonlinear regression, yielding the
following values for the empirical
parameters: n = 0.7071, a = 34.80, and
b*-1.009. The resulting model with
this drag coefficient formulation was
then able to predict 4 additional
measured trajectories and 39 addi-
tional measured trajectory endpoints
with acceptable accuracy. Thus, the
influence of droplet spacing on the
local drag coefficient of a single
droplet has been quantified.
A numerical model has been
developed to predict the ballistics of
an isolated burning droplet. This
model includes the effects of droplet
interaction on drag and evaporation
rate, and turbulence effects on
droplet penetration into the com-
bustion environment. Experiments
on a laminar flow flat-flame burner,
and a 100 kW swirling, turbulent
combustor have been conducted to
calibrate the droplet ballistics model.
Experimental results have shown that
droplet penetration increases with
increasing droplet initial size and
initial velocity and decreases with
increasing initial spacing. Model
predictions on droplet penetration
are very close to experimental
findings except for changing initial
spacing of droplets in a stream.
This Project Summary was
developed by EPA's Air and Energy
Engineering Research Laboratory,
Research Triangle Park, NC, to
announce key findings of the research
project that is fully documented In a
-------
separate report of the same title (see
Project Report ordering information at
back).
Introduction
Of the hazardous organic wastes
produced in the U.S., about 75% are
liquids or dissolved in liquids. With
landfill disposal of these wastes
becoming increasingly unpopular due to
growing public concern and increased
levels of hazardous waste production,
thermal destruction is being considered
as an alternative disposal method.
Successful incineration requires that
flame radicals and high temperatures be
combined to destroy the principal
organic hazardous components (POHCs)
of the feed waste, and to minimize the
formation of products of incomplete
combustion (PICs), which may be as
hazardous as, or even more hazardous
than, the original POHCs. Proper spray
atomization provides the necessary
dispersion of liquid fuel and waste into
the oxidizer to avoid incinerator failure by
inadequate mixing. Fuel spray nozzles
degrade with use and must be replaced
periodically. Therefore, there is a need to
understand and quantify how atomization
parameters limit liquid waste incineration
so that a sound rationale for selecting
and replacing spray nozzles can be
defined.
Four group combustion modes of a
fuel droplet cloud have been identified,
with single droplet combustion being
applicable in practice to only a very lim-
ited number of special situations. Such a
special situation, however, can arise
during the incineration of liquid haz-
ardous wastes, where droplets with large
diameters (as much as an order of
magnitude larger than the mean) con-
gregate at the outer edge of fuel spray
cones. One or more of these large, errant
droplets may individually pass through,
or bypass, the main flame zone and lead
to a failure mode in the incinerator. For
example, bypassing of as few as one
drop out of 10 million can lead to failure
to meet a destruction removal efficiency
(ORE) in excess of 99.99%, as required
by law.
Motivation for this study lay, therefore,
in the need to predict single, nonmean
droplet trajectories in a turbulent flame
zone. To this end, experiments have
been conducted (1) to determine min-
imum requirements of a model that
successfully predicts measured trajec-
tories of single monodisperse droplet
streams in three-dimensional turbulent
diffusion flames, and (2) to relate droplet
penetration distance to potential incin-
erator failure modes.
A semi-emperical numerical model
has been developed for predicting the
ballistics of burning droplet streams in
turbulent diffusion flames. An important
input to this model is proper repre-
sentation of drag coefficient on a non-
evaporating droplet in a stream.
Drag Coefficient
While much information is available on
the relationship of the drag coefficient,
CD, of an isolated sphere to Reynolds
number, Re, little is available on the
dependence of CD on Re and droplet
spacing, nondimensionalized by droplet
diameter, D, as L/D. As droplet spacing
is reduced, drag is reduced due to wake
effects. For large droplet spacings, when
droplets cease to interact, the drag
coefficient, CD™, for each one is that of
an isolated sphere, and is a function only
of Re. For the slightly distorted droplets
formed by a vibrating capillary droplet
generator, the recommended relation is:
(CD)°°(Re)
= 27Re
0)
-0.84
D1 am
L/D->°°, is to use the asymptotic
expansion formula:
[CD(Re,L/D)r n = [CD(Re,L/D)] ~ n
+ [(CDf(Re)rn
where n is a parameter to be obtained b
experiment.
Trajectory Model
A simple, numerical model has beei
developed for predicting the ballistics c
an isolated burning droplet. This mod€
includes effects of droplet interaction c
drag and evaporation rate, evaporatioi
effects on drag, and turbulence effects oi
penetration. The model is used to solvi
the uncoupled equations of drople
motion in a Lagrangian framework, i
three-dimensional grid structure is es
tablished for specifying the backgroum
gas velocity (vg), temperature (Tg), am
chemical speciation. Calculations an
terminated when the droplet exits thi
computational domain.
Experimental Methods and
Results
The objectives of this study were (1
to measure and predict the three-dim
ensional trajectories of single mono
disperse droplet streams in turbuler
diffusion flames, and (2) to study th
relation of these trajectories with drople
incineration effectiveness.
The droplet generator is a vibratin
orifice device with ancillary electronics t
facilitate droplet spacing variation. Initi;
droplet diameter and spacing wer
measured using strobe photography
Initial droplet velocity was calculate
from the vibrator frequency setting. Th
fluid tested, chosen on the basis c
conductivity (for electrostatic chargin
and deflection) and viscosity (for drople
formation), was a mixture of 80% (b
volume) Shell Oil Company fuel additiv
ASA-3, consisting mostly of xylen
(CaHio), and 20% distillate fuel oil.
To determine droplet spacing effect
on the non-evaporating droplet dra
coefficient, trajectory experiments wer
conducted in the cold flow, quiescer
environment in an observation tunne
Empirical parameters a, b, and
(Equations 2 and 3) were evaluated.
Parameters, a, b, and n wer
estimated using data from 10 full tr£
jectories. Nonlinear regression for p{
rameter estimation was accomplishe
using an algorithm based on a mu
-------
tidimensional search in parameter space
for the minimum value of a sum of
squares functional measuring deviation
from the data. The least squares
parameter estimates were: a = 34.8,
b =-1.009, and n = 0.7072.
Measured and predicted trajectory
endpoints are shown in Figure 1 for the
10 trajectories in regression and for 39
additional trajectories.
Two-dimensional laminar flow
experiments in a bench-scale flat-
flame burner were then conducted to
calibrate the model, allowing minimum
model requirements to be determined
without the complexities of three
dimensions and turbulence. Trajectories
of nonburning and burning droplets were
measured to test the model in this aero-
dynamically simple environment. These
results, shown in Figures 2, 3, and 4,
demonstrate that droplet interaction
effects on evaporation rate and evap-
oration effects on drag are important
model requirements. An important result
from these two-dimensional laminar
flow experiments was the determination
of the value of 63, the mass transfer
coefficient of Phase 3 in the three-
phase burning liquid. A value of 0.1 for
B3 made model predictions of the droplet
burnout point to be in fair agreement with
the experimental observations.
Three-dimensional turbulent diffusion
flame experiments were conducted on a
100 kW combustor. In baseline tests
without droplet injection, the combustion
gas flow field was characterized in terms
of temperature and velocity.
Droplet trajectories were measured
with the aid of high speed photography.
Droplet destruction efficiency (DE) was
determined by measuring the increased
levels of exhaust unburned hydrocarbons
(UHC or surrogate POHC) and carbon
monoxide (CO or PIC) due to droplet
injection. These emissions were mea-
sured in the stack where the combustion
gases are well mixed. Flame ionization
was used to detect UHC, and an infrared
instrument was used for CO mea-
surement. A removable water-cooled
coil for quenching the combustion gases
was inserted, resulting in a bulk gas
residence time of 0.6 s before quench-
ing.
Droplet trajectories were observed for
a variety of initial droplet stream con-
ditions, including variation in droplet size,
velocity, spacing, and injection angle.
Finally, droplet incineration was
measured with the quench coil inserted.
: The experimentally measured droplet
axial penetration and the model
predictions are given in Table 1. Closely
spaced droplets penetrated farther than
isolated droplets. Little effect of spacing
was observed for values greater than 10
diameters. Axial penetration distance
roughly doubled with a doubling of initial
velocity. Droplet penetration increased as
droplet diameter increased, as well. Little
change in axial penetration was observed
for isolated droplet injection at 0 and 45
degrees. As shown in Table 1, these
trends are also predicted by the model
when run using 63 = 0.1 obtained earlier
in two-dimensional laminar flow exper-
iments. However, the model needs to be
improved to predict droplet spacing
effects on axial penetration a little better.
Tests were conducted with the
water-cooled coil inserted downstream
of the Type-C flame to quench the
combustion gases and measure droplet
incineration. Combustion gases were
extracted from the stack, with emissions
analyzed with and without droplet
injection. Droplet DE was calculated as
the mass of carbon emitted as CO and
UHC divided by the mass of carbon
injected.
Droplet incineration results are shown
in Figure 5. The previously measured
mean axial penetration is also shown,
with dashed lines representing the range
of droplet trajectories. These data
indicate that droplet DE is related
inversely to droplet penetration, which
has been shown to depend on droplet
atomization properties. Thus, predicting
droplet ballistics may be one tool for
anticipating incinerator failure modes due
to poor atomization.
Conclusions
Large droplet penetration of the flame
zone has been observed as a function of
droplet atomization parameters in tests
with single monodisperse droplet
streams injected into turbulent diffusion
flames. The incomplete incineration of
these hydrocarbon droplets has been
approximated by measuring CO and
UHC emissions, and a strong correlation
with droplet penetrating has been
observed. The short burning distances
and relatively long trajectories prior to
ignition exhibited suggest that droplet
aerodynamics prior to ignition is of
primary importance in understanding and
predicting ORE.
-------
I
140
120
100
80
60
40
20
4
In (L°/D)
!
140
120
100
80
60
40
20
» Re° = 292, height =14.0 cm
O Re° = 136, height =7.3 cm
A Re° = 1 10. height = 7.3 cm
» Re° = 89.6, height = 8.6 cm
345
In (L°/D)
Figure 1. Droplet trajectory endpoints. Points represent measured trajectory endpoims; lines are model results. Results from the best-fit
non-linear regression of 10 full trajectories at Re° = 211, with the endpoint being the horizontal distance (range) that a droplet
travels while falling a vertical distance (height) of 11.7 cm, are shown in the top plot. Model predictions of 39 more trajectory
endpoints are shown in the bottom plot.
-------
Liquid = Water/glycerine (nonburning liquid)
o = Experimental results
— = Model predictions
D° = 237 urn
V° = 8.236 m/s
U°/D° = 2.8
Cold Flow = 10cm
N=64
N=512
Figure 2. Nonburning droplet trajectories in a laminar two-dimensional flame.
10
x (cm)
15
20
Liquid = Fuel Oil/ASA III (burning liquid)
D° = 234 urn
V° = 8.5 m/s
L°/D° = 2.8
Cold Flow = 10cm
° = Experimentally observed endpoints
25
30
64
,128
512
N=1
Figure 3. Experimental results for trajectory measurements of differently spaced droplet
streams in a laminar two-dimensional flame.
-------
10
x (cm)
15
20
25
30
Liquid = Fuel OH/ASA III (burning liquid)
Cold Flow - 10 cm
B3 = 0.1
A = Burnout of 2nd phase
D = Burnout of 3rd phase
(Trajectory endpoint)
128
64
N=1
2
Figure 4. Model predictions for trajectories of differently spaced droplet streams in a laminar
two-dimensional flame.
-------
Table 1. Measured Droplet Axial Penetration in a Three-Dimensional
Turbulent Diffusion Flame and Corresponding Model
Predictions
Input Test Condition
Experiment Axial
Penetration (cm)
Model Axial
Penetration (cm)
Nominal: D" =
V =
L°/D° =
6" =
D" = 306 pm
= 371 nm
V = 3.8 mis
= 7.4 m/s
234 tun
6.2 mis
737
30 degrees
L°ID°
6° =
= 2.54
= 4.80
= 9.70
= 77.50
= 34.00
0 degrees
45 degrees
20-45
35-60
45-70
15-25
30-60
35-75
30-65
20-55
20-45
20-40
20-45
20-50
30-55
40-64
53-68
20-42
32-65
32-59
32-59
31-59
31-59
31-59
20-55
20-55
60
50
40
% 30
c
01
Q.
20
70
100
100 200 300 400 500
Initial Diameter (pm)
(a) Size Variation
70
/OO
• • -^•^^0 -
2 4 6 8 10 §•
Initial Velocity (m/s) ^
(b) Velocity Variation
•i • • i 1700 D
In (L°/D°)
(c) Spacing Variation
t In f |-
l 75 30 45 60 75 $ "~
Injection Angle (degrees) a.
(d) Angle Variation :$.
Figure 5. Droplet destruction efficiency and axial penetration results.
------- |