We have completed a series of both vented and sealed cookoff experiments of black powder and smokeless powder in our Sandia Instrumented Thermal Ignition (SITI) apparatus at bulk densities of 1078 and 729 kg/m3, respectively. The confining aluminum cylinder was ramped from room temperature to a set point temperature and then held at the setpoint temperature until ignition. The setpoint temperatures varied between 495 to 523 K for the black powder and 401 to 412 K for the more sensitive smokeless powder. The vented experiments show a significant delay in thermal ignition, indicating that the ignition is dependent on pressure. Post experimental debris shows greater violence for our smokeless powder experiments than our black powder experiments. A simplified universal cookoff model (UCM) was calibrated using the black powder and smokeless powder SITI data and used to predict pressurization and thermal ignition. The current work presents the first calibration of the UCM with a double base propellant. This work also presents the first pressure-dependent cookoff model for black powder and smokeless powder.
We have completed a series of vented and sealed cookoff experiments of the ε-polymorph of CL-20 in our Sandia Instrumented Thermal Ignition (SITI) apparatus using both powder and pressed pellets at nominal densities of 313 ± 8 kg/m3 and 1030 ± 4 kg/m3, respectively. The boundary temperature of our aluminum confinement cylinder was ramped in 10 minutes from room temperature to a prescribed set-point temperature ranging between 448 nd 468 K and held at the set-point temperature until ignition. A universal cookoff model (UCM) has been calibrated using the ε-CL-20 SITI data to predict pressurization and thermal ignition of ε-CL-20. The ignition model was validated by using one-dimensional time-to-explosion (ODTX) ignition data from a different laboratory. We found that a thirtyfold increase in the reaction rates due to liquefaction at 520 K could explain the high temperature ODTX cookoff data. The model gives a plausible explanation of why melting is important in fast cookoff events involving CL-20. Our model also gives support to 520 K as the liquefaction point of CL-20, which has different values in the literature.
The solubility of RDX (hexahydro-1,3,5-tri-nitro-1,3,5-triazine) in TNT (2,4,6-trinitrotoluene) at elevated temperatures is required to accurately predict the response of Comp-B3 (60:40 RDX:TNT) during accidents involving fire. As the temperature increases, the TNT component melts, the RDX partially dissolves in the liquid TNT, and the remaining RDX melts (203 ∘C) as the Comp-B thermally ignites. In the current work, we used a differential scanning calorimeter (DSC) to estimate the solubility of RDX in TNT at the melting point of RDX. Most DSC measurements of Comp-B3 do not show an RDX melt endotherm. The absence of an endotherm associated with the RDX melt has been interpreted as RDX being completely dissolved in TNT before reaching the melting point. We have observed that the endotherm is not absent, but is masked by exothermic reactions occurring at these elevated temperatures. We have inhibited the exothermic reactions by venting our DSC samples and measuring the RDX melt endotherm in our Comp-B3 samples at about 203 ∘C. Using the measured heat flow associated with the RDX melt and the latent melting enthalpy of RDX, we have approximated the solubility of RDX in TNT to be roughly 50–100 g RDX per 100 g TNT. The broad range is based on corrections for exothermic reactions occurring as the RDX melts.
Thermal conductivity has been determined for a variety of energetic materials (EMs) using finite element analysis (FEA) and cookoff data from the Sandia Instrumented Thermal Ignition (SITI) experiment. Materials studied include melt-cast, pressed, and low-density explosives. The low-density explosives were either prills or powders with some experiments run at pour density (not pressed). We have compared several of our thermal conductivities with those in the literature as well as investigated contact resistance between the confining aluminum and explosive, multidimensional heat transfer effects, and uncertainty in the thermocouple bead positions. We have determined that contact resistance is minimal in the SITI experiment, the heat transfer along the midplane is one-dimensional, and that uncertainty in the thermocouple location is greatest near the heated boundary. Our values of thermal conductivity can be used with kinetic mechanisms to accurately predict thermal profiles and energy dissipation during the cookoff of explosives.
Validated models of melt cast explosives exposed to accidental fires are essential for safety analysis. In the current work, we provide several experiments that can be used to develop and validate cookoff models of melt cast explosives such as Comp-B3 composed of 60:40 wt% RDX:TNT. We present several vented and sealed experiments from 2.5 mg to 4.2 kg of Comp-B3 in several configurations. We measured pressure, spatial temperature, and ignition time. Some experiments included borescope images obtained during both vented and sealed decomposition. We observed the TNT melt, the suspension of RDX particles in the melt, bubble formation caused by RDX decomposition, and bubble-induced mixing of the suspension. The RDX suspension did not completely dissolve, even as temperatures approached ignition. Our results contrast with published measurements of RDX solubility in hot TNT that suggest RDX would be completely dissolved at these high temperatures. These different observations are attributed to sample purity. We did not observe significant movement of the two-phase mixture until decomposition gases formed bubbles. Bubble generation was inhibited in our sealed experiments and suppressed mixing.
We have completed a series of small-scale cook-off experiments of ammonium nitrate (AN) prills in our Sandia Instrumented Thermal Ignition test at nominal packing densities of about 0.8 g/cm3. We increased the boundary temperature of our aluminum confinement cylinder from room temperature to a prescribed set-point temperature in 10 min. Our set-point temperature ranged from 508 to 538 K. The external temperature of the confining cylinder was held at the set-point temperature until ignition. We used type K thermocouples to measure temperatures associated with several polymorphic phase changes as well as melting and boiling. As the AN boiled, our thermocouples were destroyed by corrosion, which may have been caused by reaction of hot nitric acid (HNO3) with nickel to form nickel nitrate, Ni(NO3)2. Videos of the corroding thermocouples showed a green solution that was similar to the color of Ni(NO3)2. We found that ignition was imminent as the AN boiling point was exceeded. Ignition of the AN prills was modeled by solving the energy equation with an energy source due to desorption of moisture and decomposition of AN to form equilibrium products. A Boussinesq approximation was used in conjunction with the momentum equation to model flow of the liquid AN. We found that the prediction of ignition was not sensitive to small perturbations in the latent enthalpies.
We have used several configurations of the Sandia Instrumented Thermal Ignition (SITI) experiment to develop a pressure-dependent, four-step ignition model for a plastic bonded explosive (PBX 9407) consisting of 94 wt.% RDX (hexahydro-1,3,5-trinitro-1,3,5-triazine), and a 6 wt.% VCTFE binder (vinyl chloride/chlorotrifluoroethylene copolymer). The four steps include desorption of water, decomposition of RDX to form equilibrium products, pressure-dependent decomposition of RDX forming equilibrium products, and decomposition of the binder to form hydrogen chloride and a nonvolatile residue (NVR). We address drying, binder decomposition, and decomposition of the RDX component from the pristine state through the melt and into ignition. We used Latin Hypercube Sampling (LHS) of the parameters to determine the sensitivity of the model to variation in the parameters. We also successfully validated the model using one-dimensional time-to-explosion (ODTX and P-ODTX) data from a different laboratory. Our SITI test matrix included 1) different densities ranging from 0.7 to 1.63 g/cm3, 2) free gas volumes ranging from 1.2 to 38 cm3, and 3) boundary temperatures ranging from 170 to 190 °C. We measured internal temperatures using embedded thermocouples at various radial locations as well as pressure using tubing that was connected from the free gas volume (ullage) to a pressure gauge. We also measured gas flow from our vented experiments. A borescope was included to obtain in situ video during some SITI experiments. We observed significant changes in the explosive volume prior to ignition. Our model, in conjunction with data observations, imply that internal accumulation of decomposition gases in high density PBX 9407 (90% of the theoretical maximum density) can contribute to significant strain whether or not the experiment is vented or sealed.
We combined the miniature cylinder expansion test (Mini-Cylex), with the Disk Acceleration Test (DAX) using Type K copper, Picatinny Liquid Explosive, and photonic Doppler velocimetry. We estimated the CJ state using plate reverberation methods at the test cap. We extracted velocities at 2, 7, and 10 volume expansions to fit Jones-Wilkins-Lee Equation of State and estimated Gurney velocity at the tube wall. The test also provides an additional method to estimate reaction products Hugoniot through knowledge of the copper test cap. Our experiments and simulations are within expected uncertainty. The test and the analysis effectively reduce costs while keeping or increasing fidelity.
In previous studies, we found that the nitroplasticizer in the HMX-based explosive PBX 9501 played a crucial role in cookoff, especially when predicting response in larger systems. We have recently completed experiments with a similar explosive, LX-14, that has a relatively nonreactive binder. We expected the ignition times for LX-14 to be longer than PBX 9501 since PBX 9501 has a more reactive binder. However, our experiments show the opposite trend. This paradox can be explained by retention of reactive gases within the interior of LX-14 by the higher strength binder resulting in faster ignition times. In contrast, the binder in PBX 9501 melts at low temperatures and does not retain decomposition gases as well as the LX- 14 binder. Retention of reactive gases in LX-14 may also explain the more violent response in oblique impact tests when compared to PBX 9501.
A diffusion-limited reaction model was calibrated for Al/Pt multilayers ignited on oxidized silicon, sapphire, and tungsten substrates, as well as for some Al/Pt multilayers ignited as free-standing foils. The model was implemented in a finite element analysis code and used to match experimental burn front velocity data collected from several years of testing at Sandia National Laboratories. Moreover, both the simulations and experiments reveal well-defined quench limits in the total Al + Pt layer (i.e., bilayer) thickness. At these limits, the heat generated from atomic diffusion is insufficient to support a self-propagating wave front on top of the substrates. Quench limits for reactive multilayers are seldom reported and are found to depend on the thermal properties of the individual layers. Here, the diffusion-limited reaction model is generalized to allow for temperature- and composition-dependent material properties, phase change, and anisotropic thermal conductivity. Utilizing this increase in model fidelity, excellent overall agreement is shown between the simulations and experimental results with a single calibrated parameter set. However, the burn front velocities of Al/Pt multilayers ignited on tungsten substrates are over-predicted. Possible sources of error are discussed and a higher activation energy (from 41.9 kJ/mol.at. to 47.5 kJ/mol.at.) is shown to bring the simulations into agreement with the velocity data observed on tungsten substrates. This higher activation energy suggests an inhibited diffusion mechanism present at lower heating rates.
In this study, we have made reasonable cookoff predictions of large-scale explosive systems by using pressure-dependent kinetics determined from small-scale experiments. Scale-up is determined by properly accounting for pressure generated from gaseous decomposition products and the volume that these reactive gases occupy, e.g. trapped within the explosive, the system, or vented. The pressure effect on the decomposition rates has been determined for different explosives by using both vented and sealed experiments at low densities. Low-density explosives are usually permeable to decomposition gases and can be used in both vented and sealed configurations to determine pressure-dependent reaction rates. In contrast, explosives that are near the theoretical maximum density (TMD) are not as permeable to decomposition gases, and pressure-dependent kinetics are difficult to determine. Ignition in explosives at high densities can be predicted by using pressure-dependent rates determined from the low-density experiments as long as gas volume changes associated with bulk thermal expansion are also considered. In the current work, cookoff of the plastic-bonded explosives PBX 9501 and PBX 9502 is reviewed and new experimental work on LX-14 is presented. Reactive gases are formed inside these heated explosives causing large internal pressures. The pressure is released differently for each of these explosives. For PBX 9501, permeability is increased and internal pressure is relieved as the nitroplasticizer melts and decomposes. Internal pressure in PBX 9502 is relieved as the material is damaged by cracks and spalling. For LX-14, internal pressure is not relieved until the explosive thermally ignites. The current paper is an extension of work presented at the 26th ICDERS symposium [1].
Al/Pt nanolaminates with a bilayer thickness (tb, width of an Al/Pt pair-layer) of 164 nm were irradiated with single laser pulses with durations of 10 ms and 0.5 ms at 189 W/cm2 and 1189 W/cm2, respectively. The time to ignition was measured for each pulse, and shorter ignition times were observed for the higher power/shorter pulse width. Videographic images of the irradiated area shortly after ignition show a non-uniform radial brightness for the longer pulse, while the shorter pulse shows uniform brightness. A diffusion-limited single step reaction mechanism was implemented in a finite element package to model the progress from reactants to products at both pulse widths. The model captures well both the observed ignition delay and qualitative observations regarding the non-uniform radial temperature.