Session EE04: High Fidelity Modeling and Simulation II

Sensitivity of predictions of critical energy of heterogeneous energetic materials to reaction kinetics models

Detonation of heterogeneous energetic materials (HEs) is initiated at hot spots. Reaction fronts propagate from such hot spots resulting in complete consumption of the surrounding material. The meso-structures of HEs such as plastic bonded explosives (PBXs) and pressed materials are replete with voids, defects and interfaces, which are sites of hot spot formation. To predict the response of HEs at the macro-scale, the meso-scale dynamics must be captured and the localization of energy that results from shock focusing and interfacial interactions must be quantified. Capturing such highly localized events that are crucial to performance prediction require a thorough understanding of the chemical reaction kinetics that govern the decomposition of solid energetic crystals to gaseous products. In the current work, multi-scale simulations are performed to investigate the uncertainties in the chemical kinetics parameters. Ensembles of high-resolution reactive void collapse simulations are performed considering the global Arrhenius parameters for pressed HMX materials to construct a meso-informed surrogate model in a high dimensional parameter space. Then macro-scale computations of shock-to-detonation (SDT) transition are performed using the meso-informed Ignition and Growth (MES-IG). The performance of the HE at the macro-scale is evaluated via the critical energy required for initiation in the Walker-Wasley/James space. The predicted critical energy envelopes are compared with experimental data. The results quantify the effects of uncertainties in the chemical kinetics parameters on the macro-scale sensitivity predictions, and provide a best estimate for the experimental data. The meso-informed surrogate model developed in this study will guide an expansion of reaction kinetics models to reliably predict macro-scale sensitivity for other HE species such as RDX, TATB etc.   

Insights from High-Order Accurate Mesoscale Simulations of Shocked Energetic Materials

Predictive modeling of the sensitivity and performance of energetic materials (EM) hinges on the accuracy of the mesoscale simulations. The thermomechanical phenomena like hotspots, shear bands, chemical reactions etc. used to characterize the sensitivity of EMs occur in the microstructure. QoIs from the mesoscale simulations are used to develop closure models to simulate the macroscale response of EMs. Therefore, to predict the macroscale response of EMs, the mesoscale dynamics must be captured accurately. The majority of numerical schemes used for computational studies aimed at predicting the sensitivity of EMs are at best nominally 2^nd order accurate and require well-resolved simulations to obtain grid independent solutions. In this work, we employ a 5^th order accurate scheme for the mesoscale simulations of energetic materials. The levelset method is used to delineate interfaces in a sharp manner and a Riemann solver is employed across the interface to maintain high order accuracy at the sharp interface. This high order technique provides exceptional resolution of the interfacial thermomechanical dynamics at the mesoscale and offers improved accuracy of the mesoscale simulations. Unprecedented resolution of the interfacial and localization (shocks, interfaces, reaction fronts, shear bands) dynamics is revealed. This work also evaluates the balance between computational cost and accuracy of the solutions obtained with high-order accuracy methods and allows for the assessment of the accuracy of state-of-the-art numerical solutions of shock interactions with heterogeneous energetic materials.   

Crossing Species: meta-learning the shock response of CHNO energetic materials in a predictive framework

Predicting the shock response of CHNO energetic materials is critical for their safe and effective use in applications, propellants, pyrotechnics, and explosives. However, accurately and consistently predicting the shock response of CHNO is challenging since the group includes a variety of material species, and each behaves differently under the application of shock. In this study, we developed a physics-informed – “meta-learning” - (PIML) method which can learn the generalized shock response knowledge across different CHNO species and quickly adapt to make predictions across different species with a minimal amount of training data. We, present a chemical decomposition model for energetic materials, which aims to understand and predict the thermal decomposition behavior of these materials that utilize a combination of thermodynamic and kinetic data to predict the decomposition pathways and rates of these materials. Importantly, the kinetic data from this experiment can be interpreted as the sensitivity of those materials, which allowed us to develop a kinetic equation for a single step from a multistep. The kinetic parameters are obtained from solving a system of coupled differential equations that describes the time-dependent evolution of the concentrations of the species involved in the decomposition process. Later on, reactive void collapse experiments were conducted on HMX, RDX, TATB, PETN, and TNT for different pressures and diameters of the void. Consequently, we used PIML to quickly adapt the void collapse knowledge from the HMX dataset to other species. Out validation with RDX showed that PIML can predict the shock response of EM well, demonstrating agreement between results from numerical and machine learning prediction, despite being trained with a small dataset. The validated meta-learning model is then used to provide valuable insight into the underlying mechanisms of sensitivity prediction of energetic materials and could be useful for the development of new and improved materials.   

Application of machine learning to study the effect of damage on sensitivity of energetic materials at the meso-scale

Damage in energetic material (EM) microstructures can impact performance in—or even cause failure of—devices critical to national security and safety. Therefore, it is important to be able to model microstructural damage and to study the effect of loading on the response of the material at various levels of damage. Here, we present a framework called HEDS (Heterogeneous Energetic Damage Simulator), a tool to generate varied levels of damage in microstructure images of one type of plastic bonded explosive (PBX). The workflow in HEDS starts with preprocessing and importing scanned cross-sectional images of the PBXs. HEDS uses deep learning to identify areas of damage in existing images of PBX, remove them from the image, and allows the user to adjust the extent (volume fraction) of damage in the same microstructure. This architecture consists of two separate U-Nets to perform the tasks of semantic segmentation (classification and extraction of damage from PBX images) and image inpainting (generating images of pristine microstructures with no damage). HEDS stores a library of extracted damage patterns that it draws from the segmentation, which can be enriched by the user through various affine transformations of the damage pattern. When damage needs to be added back into the microstructure, the patterns of damage are drawn from this “damage library”. By progressively adding back damage into the undamaged (in-painted) microstructure comparison of the sensitivity to shock loading of PBX with varying degrees of damage is performed. Finally, we present results of shock simulations from damaged microstructures generated by HEDS.   

A Morphologically Aware Model for High Explosives

Predicting the performance and safety of explosive devices relies upon an understanding of the underlying hot spot mechanisms. Explosive compositions which only differ in microstructure are known to have significant variations in initiability. Conventional reactive flow models do not directly incorporate microstructure information, therefore different parameter sets must be developed to account for lot-to-lot variations. We develop a morphologically aware detonation model for high explosives that incorporates pore size distribution data. Pore size data is used to define the number of hot spots that are ignited as a function of the effective plastic strain. The ignition sites then burn spherically. Initial burn products react through a pseudo-diffusion-controlled reaction to form the final products. Our model is simulated using ALE3D and the parameters controlling the initiation and burning are optimized using Pop-Plot and embedded pressure gauge data from nominal experiments. The calibrated model is then used to predict the effect of altering the initial porosity distribution on the run-to-detonation.   

Inert Melting Curve and Thermal Melting Kinetics of β-HMX for 10 GPa ≤ P ≤ 40 GPa

Non-reactive all-atom molecular dynamics simulations of solid-liquid phase coexistence were used to determine the pressure-dependent melting curve T_melt (P) and kinetics of melting for (010)-oriented β-HMX on the interval 10 GPa ≤ P ≤ 40 GPa. The study extends the melting curve for  P ≤ 5 GPa due to Kroonblawd and Austin (K-A) [Mech. Mater. 152, 103644 (2021)] to detonation pressures. A time dependent, layer-by-layer analysis of molecular displacements and 2D radial distribution functions is developed to track advancement of the melt front in quasi-1D phase coexistence simulations. The melting temperature is predicted to increase from 860 K at 1 GPa to ≈ 2280 K at 40 GPa and is found to follow the empirical Simon-Glatzel (S-G) function. Assessment of the extrapolation error when only low-P data are used in the S-G fit reveals that such extrapolations are fraught. Thermal melting rate coefficients on a given isobar were extracted (i) by fitting the Arrhenius rate-law form to overall crystal-zone melting times and (ii) by fitting to first-order kinetics for disintegration of individual layers in the crystal as melting advances at a given T and P. The melting rates increase exponentially with T on a given isobar but decrease significantly with increasing P. Comparisons of the predicted melting rates to characteristic time scales for pore collapse suggest that melting is slow and thus melting kinetics should be incorporated into future, high-fidelity mesoscale simulations that explicitly resolve hot spot formation and evolution.