JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec.

3-7, 2012)
SAND2012-10207C

COOKOFF OF A MELT-CASTABLE EXPLOSIVE (COMP-B)

M.L. Hobbs, M. J. Kaneshige, and M. U. Anderson

†
Sandia National Laboratories
Albuquerque, NM

ABSTRACT

A model for cookoff of Composition B (Comp-B) was developed using data from multiple sources.
Comp-B is a mixture of 60 wt% RDX (hexahydro-1,3,5-trinitro-s-triazine) and 40 wt% TNT (2,4,6-
trinitrotoluene) with or without a wax desensitizer. The single-phase reactive flow model assumes
that the Comp-B, gas products, and condensed products move at the same velocity and are at
the same temperature. Thus, three species continuity equations, a single energy equation, and a
single laminar flow equation were solved for the three-dimensional distribution of temperature,
velocity, and species. For these low Mach number flows, the pressure is only a function of time,
i.e. P (x,y,z,t) = P(t). A simplified one-step, first-order, pressure-dependent reaction mechanism
was used as a source in the species and energy equations. A hydrostatic Boussinesq source (g)
was used for the momentum equation to determine the buoyancy driven flow. Both open
(constant pressure) and closed (dynamic pressure) systems were evaluated. Differential
Scanning Calorimeter (DSC) data indicate that the TNT component in Comp-B melts at the same
temperature as pure TNT. However, the RDX seems to dissolve in the hot TNT showing a slight
endothermic reaction at temperatures well below the melting point of RDX. Results support the
hypothesis that ignition time and ignition location during cookoff of Comp-B are strongly
dependent on reactive flow.

INTRODUCTION

Composition B (Comp-B) explosives consist of mixtures of RDX (1,3,5-Trinitro-1,3,5-

triazacyclohexane), TNT (2,4,6-Trinitrotoluene), and a desensitizing wax. In the current work,
Comp-B is assumed to be composed of 60/40 RDX/TNT by weight. Comp-B was developed prior
to WWI by the Germans and the British, and has been used in mortar shells, torpedoes,
demolition charges, warheads, shaped charges, and large bombs [1]. Comp-B is prepared by
melting TNT in a steam-jacketed kettle, adding wet RDX slowly, heating and stirring until the
water is evaporated. The Comp-B is then cast into the desired shape and cooled. Comp-B is easy
to process, and has a high detonation pressure, but fails many insensitive munitions (IM)
requirements [2]. Currently, IMX-104 is being evaluated as a replacement for Comp-B [3].
Comp-B does not pass slow and fast cookoff IM tests [2]. Consequently, the response of
Comp-B during an accident, such as a fire, is important for safety analysis. The response is
complicated by many factors such as melting of TNT, dissolution of RDX in hot liquid TNT,
pressure-dependent kinetics, buoyancy driven convective flow, forced convection caused by
decomposition gas bubbles, non-Newtonian viscosity, volumetric expansion, thermal conductivity,
specific heat, and decomposition product composition at elevated temperature and pressure.
Most modeling of runaway ignition behavior in Comp-B has focused on thermal-chemistry
effects with little regard to mechanical behavior such as pressurization and flow. For example,
Zerkle presented a decomposition model of Comp-B [4] by combining a TNT and RDX

†
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation,
a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National
Nuclear Security Administration under contract DE-AC04-94AL85000. SAND NO. 2012-XXXX.

Statement A: Approved for public release; distribution is unlimited.

decomposition mechanisms. The TNT model was based on McGuire and Tarver’s TNT
decomposition mechanism [5], and the RDX decomposition model was based on the work of
Maharrey and Behrens [6]. Each of the separate TNT and RDX decomposition models were fit to
match the ODTX data. However, when these models were combined using simple mixture rules,
the predicted ignition times for Comp-B were much higher than the measured ignition times.
Zerkle suggested the dissolution of RDX into molten TNT effectively reduces the activation
barrier, causing the mixture to ignite sooner.
The TNT in Comp-B melts ~81°C, which is the same melting point as pure TNT. When all
of the TNT is melted, the RDX remains suspended in the liquid TNT. As the temperature
increases, the RDX dissolves in the hot TNT, between 127°C and 187°C. This is lower than the
melting point of RDX, 205°C. There is little movement in the suspension until most of the RDX is
dissolved. As temperature increases, the liquid decomposes into gases that form bubbles causing
forced convection. Eventually, the reaction energy is generated faster that can be dissipated by
convection and conduction leading to runaway ignition of the hot liquid.
Decomposition of RDX and TNT decomposition have previously been modeled using a
single-step, first-order, pressure-dependent reaction with a normally distributed activation energy
[7,8]. In the current work, the same kinetic form as described in [7,8] was used to model Comp-B,
and a hydrostatic Boussinesq source (g) was used with the momentum equation to determine
the buoyancy driven flow, similar to that used in [7]. However, in [7], the melting front of TNT was
not determined. In the current work, the melting front is calculated using a temperature dependent
viscosity that is high for the solid and low for the liquid. Kinetic parameters were obtained from
data in Sandia’s instrumented thermal ignition experiment (SITI) and the model is validated using
a larger cookoff test performed in a convective oven.

EXPERIMENTAL OBSERVATIONS

TGA/DTA/DSC

Figure 1 shows data from

thermogravimetric analysis (TGA) and
differential thermal analysis (DTA) for
RDX, TNT, and Comp-B from [9]. The TGA
data, which shows mass loss with
increasing temperature, are for nominally
10 mg samples heated at ~10°C/min in a
nitrogen atmosphere. The onset of mass
loss for the RDX, TNT, and Comp-B
samples occurs at 226°C, 190°C, and
145°C, respectively. The early onset of
Comp-B decomposition may be attributed
to the hot liquid TNT dissolving the RDX
suspension. RDX reacts faster in the liquid
phase than in the solid phase as discussed
by Brill et al. [10].
The DTA results in Figure 1 show
that the TNT in Comp-B melts at the same
temperature as in pure TNT. The melting Fig. 1. DTA and TGA for RDX, TNT,
of TNT is an endothermic phase change and Comp-B (data interpolated from
that occurs at temperatures that are lower [9]).
than reaction thresholds that initiate mass loss. In contrast, the RDX melting point is absent in the
Comp-B DTA data. The Comp-B DTA curve shows the onset of an exothermic reaction at 145°C,
which is quenched at 175°C by an apparent endothermic reaction that may be related to the
dissolution of the RDX suspension.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

The trends shown in Figure 1 are not quantitative. For example, the DTA for TNT shows
an exotherm in the TNT data peaking between 279°C and 295°C. However, the TGA data for
TNT shows no reaction after 278°C. The TNT and RDX samples in Figure 1 were probably not
the same TNT and RDX used to make the Comp-B samples in Figure 1. Furthermore, the TGA
mass loss was measured in a separate experiment and may not correspond to the DTA data in
Figure 1. Combined TGA and differential scanning calorimetry (DSC) better quantify the thermal
stability of Comp-B as shown in Figure 2.

Fig. 2. Simultaneous TGA/DSC data showing effective capacitance model.

 Figures 2.A and 2.B show simultaneous TGA/DSC data for TNT, RDX, and Comp-B with
the 4.68±0.31 mg samples ramped at 10°C/min. A custom 60/40 mixture of TNT and RDX was
made and is labeled 60/40 RDX/TNT in Figures 2.B and 2.B. The custom 60/40 mixture was
made with the same lot of material that was used for the individual TNT and RDX TGA/DSC runs.
There were four TNT runs (two with common TNT flakes and two from pure four times
recrystallized TNT), two RDX runs (type II, class 1 RDX), two Comp-B runs (Comp-B3 which
does not contain wax), and four custom 60/40 RDX/TNT runs. The RDX in the custom 60/40
RDX/TNT mixture was type II class 1 RDX. Two of the custom 60/40 RDX/TNT mixtures used
common TNT flakes and the other two custom mixtures used the pure TNT. The symbols in
Figure 4.A are the remaining solid fraction calculated with the JCZS-EOS [11] at 360°C for TNT
(yellow square), Comp-B (orange triangle), the custom 60/40 RDX/TNT mixture (small green
triangle in the middle of the orange triangle), and the RDX (cyan circle).
Figure 2.C shows the six baseline corrected DSC runs (2 of the Comp-B3 runs and 4 of
the custom 60/40 RDX/TNT runs). The average of the six runs is also shown as a red line in
Figure 2.C. The integral of the average heat flow between 77°C (350 K) and 127°C (400 K) is the
latent enthalpy required to melt the TNT (-32 J/gComp-B) and is close to the value calculated from
the latent enthalpy of TNT from [12] corrected for the amount of TNT in Comp-B (e.g. 98
J/gTNT0.4 gTNT/gcomp-B = 39 J/gComp-B). The integral of the average heat flow between 127°C (400
K) and 187°C (460 K) is the enthalpy required to dissolve the RDX (-19 J/gComp-B). This is much
less than the enthalpy associated with the RDX melt (e.g. 148.5 J/gRDX0.6 gTNT/gcomp-B = 89
J/gComp-B) and is spread out over a larger temperature range.
The latent enthalpy is modeled using the effective capacitance method that is shown in
Figure 2.D and E. A trapezoidal integration of the overall energy flow is separated into latent and
caloric contributions to determine the effective heat capacitance as the TNT component in Comp-
B melts between 77°C (350 K) and 127°C (400 K) and dissolves the RDX suspension between
127°C (400 K) and 187°C (460 K).The baseline specific heat was determined from [5] using a
weighted average of the specific heat of TNT and RDX, Cp = 2.1538T + 413.15, with Cp in J/kg-
K and T in K).

SANDIA’S INSTRUMENTED THERMAL IGNITION (SITI) EXPERIMENT

Kaneshige et al. [13] described the SITI experiment which is shown schematically in
Figure 3.A. The three-dimensional mesh used to model the SITI experiment is shown in Figure
3.B. The location of the thermocouple junctions and typical measured temperatures in the center
of the 2.54 cm diameter by 2.54 cm tall cylinder are shown in Figure 3.C and 3.D, respectively.
The outside temperature of the confining aluminum cylinders is maintained at a controlled
set point. Typically, the outside temperature of the aluminum confinement was ramped from room
temperature until the set point temperature was reached. The ramp rate was controlled so that
each experiment reached the set point temperature in about 10 minutes. The experiment has a
pressure tap to monitor the pressure during the experiment. The measured pressure for run #420
is shown in Figure 3.D.
Table 1 gives the set point temperature and ignition time for eight SITI experiments. The
average mass of the Comp-B in these eight experiments was 21.97 g. The 2.54 cm (1 inch) wide
3 3
by 2.54 cm (1 inch) tall pellets have a volume of 12.87 cm giving a pellet density of 1.707 g/cm .
The average height of the 2.22 cm (0.8725 inch) wide expansion volume was 0.32 cm (0.126
3
inch) making the total volume of the expansion gaps 2.47 cm . The average bulk density of
Comp-B occupying the pellet volume and the expansion volume is 1.432 g/cc. This density was
chosen as the initial density since this volume will be occupied by Comp-B once it melts and
thermally expands.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

Table 1. SITI runs with Comp-B

Run # Tset point, K 1000/T tignition, min. Mass, g Vpellet, cm3 Vexpansion, cm3 pellet, g/cc bo, g/cc* Vented
69 453.5 2.205 22.45 21.98 12.87 2.49 1.708 1.431 N
420 451.4 2.215 36.12 21.97 12.87 2.45 1.707 1.434 N
67 448.4 2.230 46.68 22.00 12.87 2.50 1.709 1.431 N
72 445.4 2.245 92.80 21.95 12.87 2.44 1.705 1.434 N
76 443.5 2.255 153.61 21.95 12.87 2.41 1.705 1.436 N
75 441.4 2.266 201.01 22.03 12.87 2.51 1.712 1.433 N
70 453.5 2.205 198.36 21.95 12.87 2.47 1.705 1.431 Y
71 452.4 2.210 221.78 21.97 12.87 2.50 1.707 1.430 Y
average: 21.97 2.47 1.707 1.432
*bo is the mass of the pellet divided by the pellet and expansion volumes.

Figure 3 SITI A) schematic B) mesh, C) thermocouple locations, and D) typical measurements.

OVEN TEST

Figures 4.A, 4.B, and 4.C show a schematic of an oven test wherein flaked Comp-B is
melted and cooked in a convection oven. Figure 4.D shows the finite element mesh. Two 4 inch
(10 cm) diameter by 3 inch (7.6 cm) tall cylinders are shown welded together in Figure 4.E. The
bottom cylinder has a notch on the top, where it is welded to the top cylinder that has a similar
notch on both the top and the bottom. Both cylinders are filled with flaked Comp-B such that when
the Comp-B melts, the bottom cylinder fills with molten Comp-B. Three thermocouple probes are
located in the bottom cylinder and are 0.75 inches (1.9 cm) from the center at 90°, 180°, and
270°. The thermocouples are positioned at different axial locations with the top, center, and
bottom thermocouple located 1 inch (2.54 cm), 1.5 inches (3.8 cm), and 2 inches (5.1 cm) from
the back wall, respectively.
Figure 4 Schematic (A-C), picture (D) and measured temperatures (E) for oven test.
Figure 4.E shows a picture of the oven test. Figure 4.F is a plot of the measured
temperatures of the heating element, the metal sheet placed under the test apparatus, and the
oven air. The temperature of the front, back and side of the bottom cylinder show a uniform
boundary temperature on the bottom cylinder. The interior thermocouples show the melting of the
TNT at about 80°C and the subsequent liquefaction of the RDX suspension between 140 and
160°C. When the suspension of RDX particles is melted, the convective cells cause the liquid to
mix and the temperatures to pinch together. Eventually, the liquid goes into thermal runaway.

MODEL

The finite element model ARIA [14] was used to solve the one temperature mixture model
with a volumetric reaction source as shown in Table 2. All of the wetted surfaces were assumed
to have a no-slip boundary condition with the velocity set to zero at these locations. The
temperature of the gas and condensed phases are assumed to be the same, that is, Tc = Tg =
T(x,y,z,t) and the velocity of the gas phase and condensed phase are also assumed to be equal
Vc = Vg = V(x,y,z,t). The model uses a low Mach flow assumption, where the gas velocity is much
less than sound speeds and the pressure within the system is only a function of time, that is,
P(x,y,z,t) = P(t). The reaction mechanism and auxiliary equations are given in Table 3. Model
parameters are given in Table 4.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

Table 2. One temperature mixture model for low Mach number flows (P  0).
*

znRT
Gas pressure (low Mach) Pg ( x, y, z, t )  Pg (t )  (1)
Vg
dv
Momentum (integral)    v  v  P  2 v     o  gh (2)
dt
T
Energy (field) Cp   C p v  T    (k T )  qr (3)
t
* C p ,k, , , P, Pg, q, r, ,R, t, T, T , V, Vg, v , x, y, z, and z represent the bulk specific heat, thermal conductivity, region
containing Comp-B, bulk viscosity, mixture pressure, gas pressure, reaction enthalpy, reaction rate, bulk density, gas
constant, time, temperature, average temperature, volume, gas volume, bulk velocity, x-coordinate, y-coordinate, z-
coordinate, and gas compressibility, respectively.

Table 3. Reaction mechanism and auxiliary equations*

Mechanism† Comp-B (C4.58H5.61N4.82O6 )  2.41N2  2.24H2O  1.88CO2  0.25CH4  0.064H2  2.45C  5.82  106 J/kg (1)
   exp    E   E  / RT  [compb]
np
Overall Reaction Rate rA P
Po
(2)
Comp-B reaction rate rCompB  [compb]  r
d
dt
(3)
Gas reaction rate rgas  dtd [gas]  6.845r (4)
Carbon reaction rate rcarbon  [carbon]  2.450r
d
dt
(5)

  (T  Td )  

Liquid rate multiplier     (1   )15 , where   0.5 1  tanh   (6)

  wd   
Distribution parameter   invsnorm(1  [compb] / [compbo ]) (7)
Gas volume fraction   1   S f co 1  o  / c  (8)
Reacted solid fraction S f  [Comp-B]M wCompB  [carbon]M wcarbom  / bo (9)
Initial gas volume fraction o   co  bo  /  co   go  (10)
Gas volume Vg   dV (11)


 n   0.0105  Covol 
Compressibility z  1  X exp(0.298 X ) , where X      (12)
Vg    T  6620 
Moles of gas n    gas  dV (13)


Average gas temperature T   bC pTdV /  bC p dV (14)

 

Condensed density c  co 1   T  To  (15)

Vol. expansion coefficient   o  o   1T  300 / 200 (16)
* A,  carbon, compb, compbo, Covol, Cp, , E, , o, gas ,, invsnorm,   MwCompB, Mwcarbon, n, np, P, Po, r, rcarbon, rcompb, rgas, R, c,
 co ,b,  bo , go, E, Td, T, To, T , V, Vg, X, z ,  represent prefactor, volumetric thermal expansion coefficient, molar
composition of carbon, molar composition of Comp-B, initial molar composition of Compb-B, BKWS gas covolume, bulk
specific heat, delta function (0 when T < Td, 1 when T  Td), activation energy, gas volume fraction, initial gas volume
fraction, molar composition of gas, volumetric expansion factor (e.g. multiplier at 500 K), inverse of the standard normal
distribution, liquid rate multiplier, region containing Comp-B, moles of gas, pressure exponent, pressure, initial pressure,
reaction rate, carbon reaction rate, Comp-B reaction rate, gas reaction rate, gas constant, condensed density, initial
condensed density, bulk density, initial bulk density, initial gas density, standard deviation of activation energy, dissolution
temperature, temperature, initial temperature, energy averaged gas temperature, volume, gas volume, BKWS parameter,
BKWS gas compressibility, distribution parameter, respectively.
†
Confined decomposition with prolonged contact of reactive species may justify the assumption that the decomposition
products go all the way to final equilibrium products, which is certainly not observed in unconfined experiments [6]. If this
assumption is wrong, than other parts of the model must also be wrong to get a good fit to the data. For example, the heat
of reaction would be different if the products included N2O and CH2O, but the pressurization would also be different.
Table 3. Model Parameters*
Parameter Description SITI Oven Test
 Vol. expansion coefficient [12] 0.0001638 0.0001638
3
Covol BKWS covolume, Å 475 475
C p , J/kgK Bulk specific heat Figure 2.D Figure 2.D

E/R, K Activation Energy/R 22000 22000

 Initial gas volume fraction 0.18 0.03
 Expansion multiplier at 500 K 10 10
k, W/mK Thermal conductivity (T< 353 K) 0.17 0.17
k, W/mK Thermal conductivity (T> 355 K) 0.12 0.12
k, W/mK Thermal conductivity (353 ≤ T ≤ 355) linear interpolate linear interpolate
m Liquid rate multiplier 15 15

Ln A, Ln (1/s) Natural logarithm of A 35.67 35.67

, Pas Viscosity (T< 412.5 K) 0.2510
6
0.2510
6

, Pas Viscosity (T> 413.5 K) 0.2 0.2

, Pas Viscosity (412.5 K ≤ T ≤ 413.5 K) linear interpolate linear interpolate
Mwcompb, g/mol Molecular weight of Comp-B 224.1 224.1
M wg , g/mol Average gas molecular weight 28.4 28.4

M wgo , g/mol Initial gas molecular weight (air) 28.0 28.0

Po, bars (psi) Initial pressure 0.83 (12.09) 0.83 (12.09)

np Pressure exponent 0 (vented) 0.725 (confined) 0 (vented)
6 6
q, J/kg Reaction enthalpy 5.8210 5.8210
 bo , kg/m3 Initial bulk density 1432 1700

co , kg/m3 Initial condensed density 1750 1750

 go , kg/m3 Initial gas density 0.944 0.944

E / R , K Distribution parameter -500 -500

Td °C Dissolution point 453.7 453.7

Venc, cm3 Enclosure volume 0.56-0.64 0.56-0.64
3
Vrdx, cm Bulk RDX volume 12.9 12.9
3
Vtube, cm Pressure tubing volume 0.84 0
wd °C Dissolution temperature width 5 5

* The bulk specific heat includes an effective capacitance model as shown in Figure 2.D. The effective capacitance model
is used for the latent phase change of TNT and the dissolution of RDX. The thermal conductivity changes at the melting
point of TNT. The viscosity changes from a solid like material to a liquid as the RDX suspension begins to dissolve.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

The bulk densities in Table 3 for both the SITI and oven test are not the starting bulk
density in either experiment. The average initial bulk density in the SITI experiment is the density
of the pellets, pellet = 1701 kg/m , as given earlier in Table 1. In Table 3, the bulk density was
3

taken to be the mass of Comp-B divided by the volume that the pellet plus the ullage volume. The
ullage volume allows for thermal expansion of the Comp-B. In the oven test, the actual initial bulk
density is half of what is listed in Table 3, since two volumes are initially filled with Comp-B flakes.
As the Comp-B melts and thermally expands, the lower cylinder is filled with all of the material
giving the density as given in Table 3. Since the melt, flow, and filling of the lower cylinder is
beyond the scope of the current analysis, the initial bulk density was taken to be the measured
mass of Comp-B divided by the volume of the lower cylinder, as given in Table 3. The thermal
conductivity was fit to match measure temperature profiles and may be higher than given in Table
3 because of the assumption regarding the initial density. Future work should address the
discrepancy in the assumed bulk densities and the actual bulk densities.
Not all of the required properties are known for Comp-B, especially at elevated
temperatures. For example, the viscosity as a function of temperature is not known over the
entire temperature range of the experiments. In reference [1], the viscosity of Comp-B is given as
0.31 at 83°C and 0.27 at 90°C. Nunez et al. [15] found the viscosity of Comp-B to be shear
thinning with the viscosity being 0.2 at 135°C. The viscosity is complex in that it changes with the
extent of reaction. In the current work, we assume that the viscosity changes from a solid-like
material to a liquid with a viscosity of 0.2 at 140°C, which is the temperature where the RDX
suspension begins to dissolve in the hot TNT.
Both the pressure dependent reactions and the swelling characteristics of the Comp-B
are not known. This leads to model parameters that are confounded. In other words, the pressure
dependency is correlated with the available gas volume, which depends on the unknown
expansion properties of Comp-B. The volumetric expansion coefficient given in Table 1 was only
given at room temperature. This coefficient is assumed to be temperature dependent, similar to
the volumetric expansion coefficient for RDX [7]. The temperature dependency given in equation
(16) in Table 3 was chosen to give a linear temperature dependency with gamma representing
the thermal expansion multiplier at 500 K. For example,  is 10 times higher at 500 K than at 300
K when  is 10. If the swelling model is wrong, then the pressure dependency is also wrong. With
more data, the swelling and pressure dependency should be reevaluated.

RESULTS

This section compares predicted and measured temperatures, pressures, and ignition
times for the SITI experiments. As a validation, predictions of the temperature in the oven test are
also compared with data.

SITI COMPARISONS

Figure 5.A shows the measured and predicted time-to-ignition for the eight SITI
experiments listed in Table 1. The ignition plot is annotated with the run number next to each of
the data symbols. The solid symbols represent runs that were sealed and the open symbols
represent runs that are vented. For the vented experiments, the pressure exponent, np, was set to
zero. For the sealed experiments, pressure exponent was 0.725 as given in Table 3.
Figures 5-B through 5.I show the measured and predicted temperatures and pressures
for the eight SITI experiments. The location of the nine of the ten temperatures shown in Figure 5
are shown in Figure 3.C. The hottest temperature is the control temperature, which is located on
the exterior of the SITI apparatus. Experiments #70 and #71 did not have pressure
measurements since these were vented experiments. The pressure measurement was missing
from experiment #72, which was a sealed experiment. In Experiment #66 the pressure tubing
plugged. Also in experiments #75 and #76, the pressure transducer did not record pressures
greater than 5000 psig (345 bars).
The predicted time-to-ignition for the sealed experiments match the measured ignition
times adequately. However, the vented experiment #71 ignited faster than predicted. If the vent
hole clogged in experiment #71, the measured ignition time would have been faster than the
prediction, since the reactions are pressure dependent. Future work should include more vented
experiments for a better statistical representation of venting effects.
The measured temperatures are adequately described by the model up until about 430 K.
At 430 K, the model predicts that all of the Comp-B is melted and convective heat transfer causes
the ten internal temperatures to come together. This temperature equilibrium is not as rapid in the
experiments implying that the change in viscosity is not as rapid as assumed in the model.

Figure 5. Predicted and measured time-to-ignition, temperature, and pressure for the eight SITI
experiments listed in Table 1. The location of the 9 internal temperatures is shown in Figure 3.A.
The hottest temperature is the control temperature located on the exterior of the SITI apparatus.

OVEN TEST COMPARISONS

Figure 6 shows several center slices of the oven test at various times. The temperature
scale is adjusted so that the minimum temperature is blue and the maximum temperature is red.
The minimum and maximum temperatures are shown at the bottom of the cut plane on each
image. The minimum and maximum Comp-B velocity is also shown on each image.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

In Figure 6, the outer confining can is shown as a transparent grey. The solid Comp-B is
shown as a transparent surface within the transparent grey confining cylinder. At 50 minutes all of
the Comp-B is solid and thus, no inner transparent surface is shown. At 100 minutes, the outer
edges are liquid and there is a large solid plug shown as a transparent orange surface at 100
minutes. The solid Comp-B is shown as a transparent yellow surface at 150 minutes and a
greenish-blue transparent surface at times greater than 150 minutes.
The RDX in the Comp-B initially melts near the outer confining shell and moves toward
the center of the bottom thermocouple. The solid plug eventually gets smaller and starts to fall
toward the bottom of the can, before it completely melts. The maximum velocity occurs as the
final solid plug of Comp-B settles to the bottom of the apparatus and eventually melts. After the
Comp-B is completely melted, the liquid heats up and eventually self-heats, and ignites at the top
of the can.

Figure 6. Predicted temperatures and velocities of Comp-B heated in a convective oven. The
outer boundary temperature was specified using the average thermocouple measurements of the
front, back, and side of the confining can as given in Figure 4.F. The outer transparent gray
surface is the confining vessel and the inner transparent surface defines the solid Comp-B. The
temperatures listed in each plot represent the minimum (blue) and maximum (red) predicted
temperatures.
 Figure 7 shows the top, middle, and bottom probe temperatures. The approximate range
of the TNT melt and RDX dissolution are also shown in Figure 7. These ranges are based on the
DSC data shown previously in Figure 2.B. The six color plots in Figure 7 show the temperature at
a center cut plane through the experimental apparatus. The black line in the color plots at 100,
150, and 200 minutes is the 140°C isotherm which defines where the viscosity changes from a
5
high value (2.510 Pas) to a low value (0.2 Pas). The predicted ignition point is at
approximately 330 minutes and the measured ignition point is at 280 minutes. The discrepancy in
the ignition time could be related to the method of melting the Comp-B flakes. In the experiment,
the flakes were melted in the combined system as shown in the picture in Figure 4.E. In the
model, the total mass of the flakes were assumed to be located in the bottom cylinder. The
agreement between the model and experiment is adequate.

Figure 7. Measured and predicted internal temperature in the oven test. The color plots are the
temperature of a slice through the center of the test apparatus. Blue corresponds to the coldest
temperature and red corresponds to the hottest temperature. The black lines on the color plots at
100, 150, and 200 minutes define the boundary of the liquid Comp-B and the solid Comp-B.
 JANNAF 27th Propulsion Systems Hazards Joint Subcommittee Meeting, Monterrey, CA (Dec. 3-7, 2012)

SUMMARY AND CONCLUSIONS

Decomposition of Comp-B is different than decomposition of the individual constituents

that comprise Comp-B: TNT and RDX. In Comp-B, TNT melts at the same temperature as in
pristine TNT; however, the RDX does not melt at the same temperature as pristine RDX. In fact,
the hot TNT seems to dissolve the RDX at lower temperatures than the RDX melting point. As
Comp-B is heated, the TNT melts and leaves the RDX in suspension. Eventually, the suspension
begins to dissolve and the liquid starts to move as convection starts to become significant.
Eventually, the hot liquid generates more energy than can be dissipated by conduction and
convection, leading to thermal runaway. The predicted location of ignition was primarily at the top
of the experiment as the hot, less-dense liquid moved to the top of the confining apparatus.
Cookoff of Comp-B is a four-phase reactive flow problem. The four phases are the solid
Comp-B, the liquid TNT, the RDX in suspension, and the decomposition gases. In the current
work, cookoff of Comp-B was modeled as a one-phase reactive flow problem wherein the solid,
liquid, and gas were assumed to be at the same local temperature and velocity. All of the
simulations were made using a three-dimensional finite element code using multiple processors
solving one energy equation, one momentum equation, and three continuity equations. A single-
step, pressure-dependent, distributed activation energy model was used to model the
decomposition of Comp-B. The model also assumes that the rate increases by a factor of 15
once the RDX is dissolved. The decomposition product hierarchy is based on the equilibrium
composition. The pressure is calculated using a low Mach flow assumption in conjunction with the
BKWS-EOS. The kinetic parameters were obtained by matching data from the SITI experiment
and the model was validated with a larger scale oven test.
The model and the data were in fair agreement. However, some simplifying assumptions
may have led to divergence between the model and data. For example, the initial density in the
oven experiment was assumed to be the total mass of Comp-B divided by a confining volume.
However, in the experiment, flaked Comp-B was melted in situ until the containing volume was
filled with the Comp-B melt. This complexity was beyond the current modeling capability.
However, the simplified assumption did provide insight into the decomposition and eventual
cookoff of unconfined Comp-B. Noteworthy simulations of the SITI experiment included both
vented and sealed experiments. The vented and sealed experiments were used to determine the
effect of pressure on the reaction rates of Comp-B.

ACKNOWLEDGMENTS

We would like to thank Craig Tarver and Tri Tran at Lawrence Livermore National
Laboratory (LLNL) for supplying the ODTX data for Comp-B, Tom Massis at Sandia National
Laboratories (SNL) for supplying the pure TNT for the TGA/DSC samples, Bob Patton (SNL) for
running the TGA/DSC experiments, Shane Snedigar (SNL) for running the SITI experiments,
Dave Zerkle at Los Alamos National Laboratory (LANL) for many discussions regarding cookoff
and viscosity of Comp-B, Mel Baer (SNL) for discussions regarding decomposition of melt-
castable explosives, Bill Erikson (SNL) and Steven Todd (SNL) for internal review, and Clint Hall,
Anthony Geller, Leanna Minier for management support.
 REFERENCES

1. Fedoroff, B. T., Sheffield, O. E., Encyclopedia of Explosives and Related Items, PATR
2700, Volume 3, U. S. Army Research and Development Command, Picatinny Arsenal,
New Jersey, USA (1966).
2. Singh, S., Jelinek, L., Samuels, P., Di Stasio, A., Zunino, L., “IMX-104 Characterization
for DoD Qualification,” 2010 Insensitive Munitions & Energetic Materials Technology
Symposium, Munich, Germany (2010).
3. Fung, V., Ervin M., Alexander, B., Patel, C, Samuels, P., “Development and manufacture
of an Insensitive Composition B Replacement Explosive IMX-104 for Mortar
Applications,” 2010 Insensitive Munitions & Energetic Materials Technology Symposium,
Munich, Germany (2010).
4. Zerkle, D. K., “Composition B Decomposition and Ignition Model,” Thirteenth International
Detonation Symposium, ONR 351-07-01, Office of Naval Research, Arlington, VA 771
(2006).
5. McGuire R.R, and Tarver, C.M., “Chemical Decomposition Models for the Thermal
Explosion of Confined HMX, TATB, RDX, and TNT Explosives,” Seventh Symposium
(International) on Detonation, NSWC MP 82-334, Naval Surface Weapon Center, White
Oak, Silver Spring, MD, 56 (1981).
6. Maharrey, S. and Behrens, R. Jr., “Thermal Decomposition of Energetic Materials. 5.
Reaction Processes of 1,3,5-Trinitrohexahydro-s-triazine Below Its Melting Point,” J.
Phys. Chem. A., 109, 11236 (2005).
th
7. Hobbs, M. L., Steyskal, M., Kaneshige, M. J., “Modeling RDX Ignition,” JANNAF 24
Propulsion Systems Hazards Meeting, Boston, MA (2008).
8. Hobbs, M. L., Kaneshige, M. J., Gilbert, D. W., Marley, S. K., and Todd, S. N., “Modeling
TNT Ignition,” J. Phys. Chem. A, 113, 10474 (2009).
9. Dobratz, B. M., “LLNL Explosives Handbook—Properties of Chemical explosives and
Explosive Stimulants,” Lawrence Livermore National Laboratory report, DE85-015961,
Livermore, CA (1981).
10. Brill T. B., Gongwer, P. E., and Williams, G. K., “Thermal Decomposition of Energetic
Materials. 66. Kinetic Compensation Effects in HMX, RDX, and NTO,” J. Phys. Chem.,
98, 12242 (1994).
11. Hobbs, M. L., Baer, M. R., and McGee, B. C., “JCZS: An Intermolecular Potential
Database for Performing Accurate Detonation and Expansion Calculations,” Propellants,
Explosives, Pyrotechnics, 24, 269, (1999).
12. Gibbs, T. R. and Popolato, A., Editors, LASL Explosive Property Data, University of
California Press, Los Angeles, CA (1980).
13. Kaneshige, M. J., Renlund, A. M., Schmitt, R. G., and Erikson, W. W., “Cook-off
Experiments for Model Validation at Sandia National Laboratories,” Twelfth International
Detonation Symposium, ONR 333-05-2, Office of Naval Research, Arlington, VA, 821
(2002).
14. Notz, P. K., Subia, S. R., Hopkins, M. M., Moffat, H. K., and Noble, D. R. ARIA Manual
Aria 1.5: User’s Manual, Sandia National Laboratories Report SAND2007-2734,
Albuquerque, NM (2007).
15. Nunez, M. P., Zerkle, D. K., and Zucker, J. M., “The rheology of molten Composition B,”
Los Alamos Report LA-UR-12-24029, Los Alamos, NM (2012).