A physics-aware deep learning model for shear band formation around collapsing pores in shocked reactive materials

TL;DR

This work addresses shock-to-detonation phenomena in energetic materials by focusing on hotspot formation and shear-band–driven energy localization under weak-to-moderate shocks. It introduces substantial improvements to the Physics-Aware Recurrent Convolutional Neural Network (PARCv2), enabling accurate, physics-informed predictions of temperature, pressure, and interfacial dynamics during pore collapse, and compares against Fourier Neural Operator and Neural ODE baselines. Across interpolation and certain extrapolation regimes, PARCv2 outperforms the alternatives, though all models struggle with fine-scale shear-band details and low-velocity dynamics, highlighting spectral-bias and boundary-condition challenges in mesh-based PIML. The study underlines the practical potential of AI-accelerated simulations for reactive materials while outlining key limitations and directions—particularly architectural innovations needed to robustly capture multi-scale physics and weak features in time-dependent, nonlinear shock problems.

Abstract

Modeling shock-to-detonation phenomena in energetic materials (EMs) requires capturing complex physical processes such as strong shocks, rapid changes in microstructural morphology, and nonlinear dynamics of chemical reaction fronts. These processes participate in energy localization at hotspots, which initiate chemical energy release leading to detonation. This study addresses the formation of hotspots in crystalline EMs subjected to weak-to-moderate shock loading, which, despite its critical relevance to the safe storage and handling of EMs, remains underexplored compared to the well-studied strong shock conditions. To overcome the computational challenges associated with direct numerical simulations, we advance the Physics-Aware Recurrent Convolutional Neural Network (PARCv2), which has been shown to be capable of predicting strong shock responses in EMs. We improved the architecture of PARCv2 to rapidly predict shear localizations and plastic heating, which play important roles in the weak-to-moderate shock regime. PARCv2 is benchmarked against two widely used physics-informed models, namely, Fourier neural operator and neural ordinary differential equation; we demonstrate its superior performance in capturing the spatiotemporal dynamics of shear band formation. While all models exhibit certain failure modes, our findings underscore the importance of domain-specific considerations in developing robust AI-accelerated simulation tools for reactive materials.

Figures (19)

• Figure 1: Simulation configuration for this work. A rectangular block of RDX with a single $50\,\mathrm{nm}$ pore impacts a rigid lower wall, generating a supported shock wave that propagates upward and initiates pore collapse and shear localization.
• Figure 2: Flowchart of this work.
• Figure 3: Initial impact velocity versus root mean squared error (RMSE) of model predictions Shaded areas indicates coverage of training set.
• Figure 4: Temperature field evolution of initial velocity of 1800 m/s.
• Figure 5: Pressure field evolution of initial velocity of 1800 m/s.