Page 1
1
CHARACTERIZATION OF ANFO USING AQUARIUM TEST
AND NUMERICAL MODELING METHODS
Eduardo Lozano1, Vilem Petr
2
AXPRO Group, Colorado School of Mines, CO, USA
Abstract
ANFO (ammonium nitrate/fuel oil) has found wide use in coal mining, quarrying,
metal mining, and civil construction in undemanding applications where the advantages of
ANFO's low cost and ease of use matter more than the benefits offered by conventional
industrial explosives. However, because of this type of explosive is highly non-ideal, its
detonation process poses great challenge. This work is focused on the development of a
model for predicting explosive power and performance from the detonation of ANFO using
an Euler-Lagrange model in Ansys Autodyn. This model was calibrated through experimental
aquarium test where the front curvature, reaction zone structure and front velocity were
measured using High Speed Imaginary. Good agreement is observed between the simulation
and experiment, which indicates that the proposed method performs well in prediction of
behavior of ANFO and having important applications in its effective use.
Keywords: High explosive, non-ideal, velocity of detonation, aquarium test, Autodyn.
I. INTRODUCTION
Non-ideal explosives are categorized as explosives that release some of their energy after the
passing of the detonation front. In this situation, the detonation zone has a non-negligible
volume and experiences interesting phenomena that may not be observed in an ideal
explosive. A common non-ideal explosive is ammonium nitrate with fuel oil (ANFO). Since
ANFO is a non-ideal explosive, studying the propagation of detonation of the ANFO can
provide insight into effects not highly known.
Numerical modeling constitutes one powerful tool in the analysis of non-ideal detonations
and provides us with unique data which is highly difficult to obtain through experimental
methods. This article presents numerical simulation of the underwater explosion of an ANFO
cylinder using a two dimensional model of ANSYS-Autodyn explicit software for nonlinear
dynamics. However, because of this type of explosive is highly non-ideal, its detonation
process poses great challenge for accurate numerical modeling.
For this reason, AXPRO Group decided to conduct experimental testing of ANFO charges
using its high fidelity detonation physics laboratory and providing valuable data in order to
1 Eduardo Lozano, AXPRO Group, Colorado School of Mines, 1600 Illinois Street, Brown Building, Room 129,
Golden, Colorado 80401, [email protected] . 2 Vilem Petr, AXPRO Group, Colorado School of Mines, 1600 Illinois Street, Brown Building, Room 120,
Golden, Colorado 80401, [email protected] .
Page 2
2
better understand the detonation process and calibrate the numerical model. This testing will
consider different parameters such as charge size and confinement due to their influence on
the detonation behavior of non-ideal explosives [1]. This study will improve our
understanding of non-ideal explosives, in general, and ANFO, in detail.
II. EXPERIMENTAL TESTING
In the experiment discussed in this paper, the detonation velocities along with the area of the
detonation zone are analyzed. This experiment was done using a cylindrical charge of ANFO
placed inside an aquarium in order to develop a detailed study about the detonation physics of
this non-ideal explosive.
A. Test Setup
The overall setup consists of a water-filled aquarium placed in the center of the test pit at the
AXPRO Explosives Research Laboratory at Idaho Springs, Colorado. The high-speed camera
used will be protected from the blast through placement in a blast-resistant hut. To avoid
damage to the camera, it will not be looking at the aquarium directly, but instead, a mirror
will be placed so that the camera can view the events indirectly.
The main ANFO charge was created on-site the day of the test for avoiding transportation
risks and storing time. For this charge, the minimum density desired was 0.85 g/cc. To
achieve this density, the plexiglass cylinder was measured and divided into three equal-in-
length sections to monitor the density of the charge as the cylinder was being loaded with
ANFO. To measure the density, the mass of the added ANFO was measured using a scale,
and then the volume of the charge was calculated. The weight of the cylinder was 2.75 kg,
but the scale was zeroed with the cylinder on top of it in order to avoid subtracting out the
mass of the cylinder each time. Figure 1 shows the dimensions of the plexiglass cylinder and
charge. The image from the camera has refraction effects associated with it, so the thickness
of the wall of the cylinder is enlarged.
The Exploding Bridge Wire (EBW) firing system was used for initiation of the EBW
detonator. An Austin White Cap Booster was used between the EBW and the main charge.
The inner length of the plexiglass cylinder was 917.6 mm long, but as a booster weighing
1.35 kg with a non-negligible volume was going to be added to the top of the ANFO column,
the ANFO column needed to stop before the top of the cylinder. For this reason, it is 876.3
mm long. However, since the equal sections were already measured out at 304.8 mm, the last
section needed to be stopped before the end. As long as the density and total mass were
correct, this early stoppage would not be a problem.
As the first section was being loaded, to ensure more tight packing, pressure was applied to
the ANFO. After the first section being loaded with 6.5 kg, the density was 0.87 g/cc. The
second addition was 6.4 kg, bringing the total mass to 12.9 kg and the overall density to 0.86
g/cc. This addition also received applied pressure, but not as much as the first. The last
Page 3
3
addition brought the total mass of the ANFO to 18.4 kg, only a 5.5 kg increase. However, as
the last addition did not take up as much volume as the previous two, the density remained at
0.86 g/cc. With the ANFO and the Booster loaded, the total mass of the charge was 19.75 kg;
with the cylinder, the mass was 22.5 kg.
TABLE I. Measured values of the ANFO charge
shows the measured values of the charge.
FIG 1. Cross-section drawing of plexiglass cylinder and ANFO column
TABLE I. Measured values of the ANFO charge
Weight of
ANFO
Volume of
Cylinder
Volume of
ANFO
Weight of
Booster
Weight of
Cylinder
Calculated
Loading
Density
18.4 kg 22380 cm3
21400 cm3
1.35 kg 2.75 kg 0.86 g/cc
Figure 2 shows the aquarium setup with the charge. The glass of the aquarium is 914.4 mm in
all directions.
Page 4
4
FIG 2. Overall setup for the aquarium for the ANFO charge
A 813 mm by 813 mm by 610 mm wooden stand supported an aquarium that measured 915
mm by 915 mm by 915 mm. The charges were placed into the aquarium as shown in Figure 1
and Figure 2. A paper grid provided one inch fiducial markings and was mounted to the back
wall of the aquarium.
Figure 3 show the layout of the test arena. In order to mitigate damage, concrete fragment
barriers were utilized. This consists of 610 mm by 610 mm by 1830 mm concrete blocks
stacked two or three high. The blocks next to the camera hut were stacked two high while the
blocks next to the aquarium were stacked three high.
Page 5
5
The SIM camera was housed in the camera hut which viewed the test stand through a turning
mirror. The dimensions go from the camera hut to the turning mirror and then from the mirror
to the center of the charge inside the tank.
FIG 3. Arena layout diagram for ANFO test
The primary camera was the Specialized Instrument (SIM) X16 camera. The objective for
this camera was to capture detonation velocity, shock wave information traveling through the
water, and case expansion information.
Because of the short exposure times, additional lighting was required for the SIM camera.
This was accomplished with the MegaSun system. Seismograph equipment was utilized with
sensors placed outside the ERL office trailer, and near the primary gate at the school house.
There were two DG645 delay generators used. The first one generated the master trigger
from time zero to T+100 µs and went from zero volts to five volts and was used to trigger the
second delay generator.
For the second delay generator, Channel AB was set to make a 1 µs wide pulse at T+10 µs.
Channels CD, EF, and GH were all set to make 1 µs wide pulses at T+100 µs. Channel AB
was used to trigger the MegaSun, Channel CD was used for the firing system, and Channel
EF was used to trigger the SIM camera. All of these channels made pulses that start at zero
volts with a pulse height of five volts, except the rear output which was fixed at 30 Volts.
Page 6
6
Scope 1 was set to record 200 µs of data while scopes 2 and 3 were set to record 400 µs of
data, sampled every 0.2 ns. Channel 1 of each scope was used to record the dual pulse signal
generated by the rear T0 output of delay generator 1. This was done to verify timing and
synchronization. This signal was also used to trigger all scopes on their AUX inputs.
Channels 2, 3, and 4 of the first scope were used to record the AB, CD, and EF outputs of the
second delay generator. Channel 2 of the second scope recorded the signal from the
photodiode which was placed in the arena to acquire the total light generated by the test
event. This light was purposely acquired indirectly to prevent the photodiode from
saturating. Channel 4 of the second scope was used to record the timing output signals of the
ultra-high-speed camera.
B. Test Results
The results shown will mainly focus on the detonation zone and the changes observed. As the
data will show, the detonation zone had a varying area and varying front and back detonation
velocities. It is important to note, however, that the areas presented in this analysis are just
the two dimensional areas as seen through the captured images. Although the areas are not
surface areas, these areas are still representative of the effects within the charge and the
detonation zone. Also, the times listed are times from the start of the detonation.
To compare the following results to a picture of what is happening, at earlier times, the
detonation is happening higher in the charge column while the later times are when the
detonation has propagated into the lower portion of the charge.
The detonation zone had a time varying area. The image data will be presented in frames, but
the analysis will depend on the absolute time of each frame. The ANFO data will begin at
100 microseconds.
Page 7
7
FIG 4. Different frames during the detonation process.
The first usable frame that had a full detonation zone was frame 4 at 125.989 microseconds.
Figure 5 shows the total area of the detonation zone for the ANFO test, and Table II shows
the measured values.
Page 8
8
FIG 5. Area of detonation zone of ANFO test as measured in pixels
TABLE II. Measured areas of detonation zone per frame in ANFO test
Time (µs) 126.0 134.7 143.3 152.0 160.6 169.3 178.0 186.6 195.3 204.0 212.6 221.3
Area (px) 13424 13988 13683 12857 12130 15717 15583 13458 17585 11912 13246 10235
Looking at the graph, the area of the detonation zone seems to vary significantly, with a
minimum of 10,235 pixels at 221.3 µs, and a maximum of 17,585 pixels at 195.3 µs. There
does not seem to be a trend in the fluctuations as they appear random.
For the following velocity data, the measurements were all first done in pixels. Three points
were chosen along the front and back of the detonation zone. Since the detonation proceeded
vertically in the images, the horizontal positions of the chosen points did not change, only the
vertical positions, allowing only the vertical velocity to be measured. Since the change in
vertical pixels and the change in time were both known from the images, the velocity was
easily calculated in pixels per microsecond (px/µs). Then, once the velocity for each of the
three points was calculated, an average was taken to find the velocity of the back and the
front of the detonation zone.
Page 9
9
At this point, the velocity was known in px/µs, but the velocity needed to be in meters per
second (m/s). This conversion was done by finding the number of pixels in one inch. The one
inch measurement came from the fiducial grid on the aquarium. Then, with this conversion,
the velocity was known in inches per microsecond. These units were easy then to convert into
m/s. Therefore, the velocity measurements are all in m/s in the following data. It is important
to note, however, that the change in position does not happen with respect to one stationary
frame but instead with respect to only the previous frames. Finding the velocity this way
makes it possible to compare the differences between the front and back velocities from
frame to frame with the area difference from frame to frame.
The velocity of the front and back of the detonation zone is important to analyze to see how
the detonation is propagating through the charge. For an ideal explosive that has a negligible
size for the detonation front, the velocity of the front and back does not vary; however, as
ANFO is a non-ideal explosive, the velocities and their changes are important to consider.
Figure 6 shows the graph of the velocities of the front and back of the detonation zone.
FIG 6. Velocities of the front and back of the detonation zone for the ANFO test
The graph shows that both the front and back velocities vary significantly as the detonation
zone progresses through the charge. The velocity of the front seems to also vary more
Page 10
10
significantly that the velocity of the back. Figure 7 examines the differences between the
front and the back velocities.
FIG 7. Difference of the velocities of the front and back of the detonation zone in the ANFO
test
The difference was calculated by subtracting the front and back velocities, so a positive
difference means that the front velocity was larger than the back velocity for that time, while
a negative difference means the back velocity was larger than the front velocity. Intuitively, if
the velocity of the front is larger than the velocity of the back of the detonation zone, the
detonation zone must increase in size.
III. NUMERICAL SIMULATION
The same experiment described before is performed using Ansys Autodyn. 18.4 Kg of
ANFO and 1.35 kg of Booster is exploded in water and detonation velocity is being measured
at different points along the ANFO charge from the center of explosion to the end of the
cylinder. The Booster used during the testing was made of Pentolite high explosive so the
same is distribution is used in the numerical model.
Page 11
11
Blast in water is one dimensional in nature. One-dimensional simulation in Autodyn can be
modelled using 2D-axisymmetric solver and with the shape of a wedge [2]. But, we use this
configuration when we are interested in the effects of an underwater explosion on the
surrounding environment. In our particular case, our objective is the study of the detonation
phenomena produced in the charge. For this reason, the experiment is setup as a 2D-
axisymmetric using the shape of a box. Furthermore, both charges (ANFO and Booster) are
modeled as Lagrange surfaces meanwhile the water is consider as Euler.
A. Material Properties
All materials are selected from Autodyn material library but ANFO parameters are calibrated
according to the previous Aquarium Test and Souer et al [3] in order to represent the behavior
of this explosive with the higher level of accuracy possible. ANFO and Pentolite have John–
Wilkins–Lee (JWL) Equation of State. The JWL EOS is the most appropriate equation used
for modelling explosives and its definition is as follows (1):
TABLE III. JWL parameters for ANFO and Pentolite.
Parameter ANFO Pentolite
Reference Density (g/cm3) 8.6E-01 1.70E+00
A (KPa) 4.95E+07 5.41E+08
B (KPa) 1.89E+06 9.37E+06
R1 3.91E+00 4.50E+00
R2 1.12E+00 1.10E+00
w 3.33E-01 3.50E-01
CJ Detonation Velocity (m/s) 4.16E+03 7.53E+03
CJ Energy/Unit volume (KJ/m3) 3.48E+06 8.10E+06
CJ Pressure (KPa) 5.15E+06 2.55E+07
The incompressibility of water in Autodyn has been modelled using Shock Equation of State
[2]. Table III shows the parameters values used by Autodyn to model the water with shock
Page 12
12
EOS. Shock EOS has Rankine-Hugoniot equations for the shock jump condition can be
regarded as defining a relation between any pair of the variables (2).
Mie-Gruneisen form of EOS based on the shock Hugoiot (3) is used:
Where,
TABLE IV. Water Shock EOS parameters.
Parameter Water
Reference Density (g/cm3) 9.98E-01
Gruneisen Coefficient 0.00E+00
C1 1.65E+03
S1 1.92E+00
S2 0.00E+00
VE/V0 0.00E+00
VB/V0 0.00E+00
C2 0.00E+00
Reference Temperature (K) 0.00E+00
Specific Heat (J/KgK) 0.00E+00
Thermal Conductivity (J/mKs) 0.00E+00
B. Model Setup
An Euler Box with the original aquarium dimensions is built and filled with water. The box
mesh also presents a grading as we move away from the location of the charge in the
aquarium. The purpose of this mesh grading is to reduce the computational effort and
resolution time.
Page 13
13
Then, three Lagrange parts are created. The first one models the Pentolite booster (red); the
second one the ANFO main charge (green); and the third one represents the Plexiglas
confinement. Additionally, fifteen theoretical gauges are placed along the charge in
equidistant points throughout the ANFO cylinder. This configuration is shown in Figure 8.
FIG 8. ANFO aquarium model set up.
A flow-out boundary condition is defined on the Aquarium sides. This condition simulates an
infinity water mass where the effects of shock wave reflections with the aquarium walls are
not considered. The detonation is initiated on the top of the Booster and the explosive
reaction is propagated along the booster and the main charge. The time limit is set to 0.3
milliseconds according to the experimental results showed in the previous section.
Lagrange-Lagrange interaction is involved in the calculation. We need to check the gap size
and gap type. If internal gap is used, the time step is usually calculated from the gap size
because it is the smallest. The internal gap size decreases with the decreasing cell length and
thus causes the smaller time step size so the best type of gap for our calculations will be
external gap.
Autodyn calculates a minimum gap size between the two parts when External gap interaction
is used. In this case, the minimum gap size is 0.1mm between the booster, main charge and
confinement, however, this distance is negligible for the simulation results.
Finally, we must prevent the erosion of degenerate cells. Some regions of the Pentolite and
ANFO will detonate outside the aquarium and those regions will undergo to large cell
Page 14
14
deformations. We are not interested in the study of those parts so we must avoid their impact
on the simulation.
C. Model Results
The fifteen gauges placed along the ANFO cylinder show the absolute velocity during the
simulation in different points. Figure 9 shows the gauge history provided by Autodyn. The
peak shown for each gauge represents the velocity of the detonation front when it reaches the
gauge. The purpose of using a large number of gauges is to obtain an accurate curve which
represents the variation of the velocity of detonation along the ANFO charge. This curve will
be compared with experimental results obtained in the Aquarium Test.
FIG 9. Velocity Gauge History of the ANFO main
Table V and Figure 10 show the velocity of the detonation front in different instants using the
peak values provide by the Autodyn Gauge History.
Page 15
15
TABLE V. Velocity of the Detonation Front in ANFO.
Gauge Time (ms) Velocity (m/s)
1 3,721 0.020774
2 3,677 0.033677
3 3,752 0.04666
4 3,696 0.059422
5 3,694 0.0724
6 3,858 0.085363
7 4,017 0.09833
8 4,016 0.11123
9 3,932 0.12419
10 3,917 0.13728
11 3,888 0.1539
12 3,898 0.16349
13 3,872 0.17642
14 3,853 0.18944
15 3,831 0.20252
FIG 10. Velocity of the Detonation Front in ANFO
3,000
3,200
3,400
3,600
3,800
4,000
4,200
4,400
0.01 0.06 0.11 0.16 0.21
Ve
loci
ty (
m/s
)
Time (ms)
ANFO Detonation Velocity
Page 16
16
The next four figures show different instants during the execution of the simulation. The
variable represented is the contour Absolute Velocity where the color gradient provides us
with graphic information about how the detonation front travels through the charge.
FIG 11. Detonation Front 0.030 milliseconds
FIG 12. Detonation Front 0.100 milliseconds
Page 17
17
FIG 13. Detonation Front 0.170 milliseconds
FIG 14. Detonation Front 0.200 milliseconds
Page 18
18
D. Numerical Model Validation
The validation of the data requires comparing the results obtained by the experimental
aquarium test and the numerical model. Figure 15 shows different values of the Velocity of
Detonation measured for the aquarium test and the Autodyn simulation. We must consider
that several physical effects, such as shock waves reflection within the aquarium or charge
package, are obviated in the numerical model. Also, the Aquarium Test just provided us with
values from the second half of the cylinder, so no experimental data is available for the first
part of the charge.
The experimental aquarium test is represented by its exponential regression line which shows
how the velocity of detonation of the ANFO charge vary until it reaches an ideal steady state.
Additionally, the velocity of detonation obtained from Autodyn is represented by the line
which merges the velocity peak points measured by the fifteen theoretical gauges. The
comparison between model and experiment is performed at 682.5mm from the leading end of
the charge. This point corresponds with 3.8 charge diameters and where is assumed that the
steady state has been reached [1].
FIG 15. ANFO Detonation Velocity Comparison between the Experimental Test and
Numerical Model
The accuracy of the numerical model versus the experimental results is measured using the
percent error. The approximation error in some data is the discrepancy between an exact
Page 19
19
value and some approximation to it [5]. Given some value vexperiment and its approximation
vmodel, the percent error (4) is as follows:
Just some specific points are available in the experimental result so it is not possible represent
the experimental velocity as a function of time with sufficient precision. Thus, the percent
error is calculated using average values from the experiment and the model. These average
values are obtained by the arithmetic mean [6] which formula (5) is as follows:
Using the available experimental data, the average detonation front velocity for ANFO
obtained in the Aquarium Test is:
According to the values of velocity measured by the gauges during the simulation, the
average detonation front velocity for ANFO is:
The percent error of the numerical model obtained using the equation (5) with average values
is:
IV. CONCLUSION
In this paper, the detonation phenomenon of an ANFO cylinder in water is successfully
simulated and the velocity of detonation along the charge is calculated using Autodyn.
Maximum values of velocity obtained by fifteen different gauges are compared with the
experimental data obtained in a previous Aquarium Test performed by AXPRO Research
Group using its high fidelity detonation physics laboratory.
The results of the experimental measurements have shown that the water-tank technique is
very useful in determining the properties of non-ideal explosives such as ANFO.
Additionally, the detonation process of this type of explosives can be successfully simulated
using a computational numerical model. Simulation studies show that Autodyn results match
very well with the theory of the detonation front traveling through the charge when the steady
state is reached.
Page 20
20
ACKNOWLEDGMENTS
The recommendations and conclusions contained in this report are the result of
a collaborative effort by members of the AXPRO Explosives Research Group. Special thanks
go to Ray Johnson, Susan Rainey, Bob Lynch, Pavlo Ivanusa and the rest of the AXPRO
team. Last, we appreciate the support provided by the Mining Engineering Department at
Colorado School of Mines for the use of the Explosives Research Laboratory at Idaho
Springs.
REFERENCES
[1] Craig B.G., Johnson J.N., Mader C.L., Lederman G.F., 1978 “Characterization of Two
Commercial Explosives”, Los Alamos Scientific Laboratory Tech. Rep. LA-7140.
[2] Neetu Jha, B. S. Kiran Kumar, 2014 “Under Water Explosion Pressure Prediction and
Validation Using ANSYS/AUTODYN”, International Journal of Science and Research
(IJSR), Volume 3, Issue 12.
[3] Souer P.C., Vitello P., 2004 “ANFO Calculations for Sedat Ensen”, Lawrence Livermore
National Laboratory, UCRL-TR-204259.
[4] Price M.A., Ghee A.H., 2009 “Modeling for Detonation and Energy Release from
Peroxides and Non-Ideal Improvised Explosives”, Central European Journal of Energetic
Materials, ISSN 1733-7178.
[5] Merigo J. M., Cananovas M., 2009 "The Generalized Hybrid Averaging Operator and its
Application in Decision Making", Journal of Quantitative Methods for Economics and
Business Administration 9: 69–84. ISSN 1886-516X.
[6] Golub G., Van Loan C.F., 1996 “Matrix Computations – Third Edition. Baltimore: The
Johns Hopkins University”, Press. p. 53. ISBN 0-8018-5413-X.
[7] Huang H., Jiao Q.J., Nie J.X., Qin J.F., 2011 ” Numerical Modeling of Underwater
Explosion by One-Dimensional ANSYS-AUTODYN”, Journal of Energetic Materials, 29:4,
292-325, DOI: 10.1080/07370652.2010.527898.