Search for: All records

We present a machine learning framework to train and validate neural networks to predict the anisotropic elastic response of a monoclinic organic molecular crystal known as ‐HMX in the geometrical nonlinear regime. A filtered molecular dynamic (MD) simulations database is used to train neural networks with a Sobolev norm that uses the stress measure and a reference configuration to deduce the elastic stored free energy functional. To improve the accuracy of the elasticity tangent predictions originating from the learned stored free energy, a transfer learning technique is used to introduce additional tangential constraints from the data while necessary conditions (e.g., strong ellipticity, crystallographic symmetry) for the correctness of the model are either introduced as additional physical constraints or incorporated in the validation tests. Assessment of the neural networks is based on (1) the accuracy with which they reproduce the bottom‐line constitutive responses predicted by MD, (2) the robustness of the models measured by detailed examination of their stability and uniqueness, and (3) the admissibility of the predicted responses with respect to mechanics principles in the finite‐deformation regime. We compare the training efficiency of the neural networks under different Sobolev constraints and assess the accuracy and robustness of the models against MD benchmarks for ‐HMX.

All‐atom molecular dynamics (MD) and Eulerian continuum simulations, performed using the LAMMPS and SCIMITAR3D codes, respectively, were used to study thermo‐mechanical aspects of the shock‐induced collapse of an initially cylindrical 50 nm diameter pore in single crystals of 1,3,5‐triamino‐2,4,6‐trinitrobenzene (TATB). Three impact speeds, 0.5 km s⁻¹, 1.0 km s⁻¹and 2.0 km s⁻¹, were used to generate the shocks. These impact conditions are expected to yield collapse mechanisms ranging from predominantly visco‐plastic to hydrodynamic. For the MD studies, three crystal orientations (relative to shock‐propagation direction) were examined that span the limiting cases with respect to the crystalline anisotropy in TATB. An isotropic constitutive model was used for the continuum simulations, thus crystal anisotropy is absent. The evolution of spatiotemporally resolved quantities during collapse is reported including local pressure, temperature, pore size and shape, and material flow. Multiple models for the melting curve and specific heat were studied. Within the isotropic elastic/perfectly plastic continuum framework and for the range of impact conditions studied, the specific heat and melting curve sub‐models are shown to have modest effects on the continuum hotspot predictions in the present inert calculation. Treating the MD predictions as ‘ground truth’, albeit with a classical rather than quantum‐like heat capacity, it is clear that – at a minimum – an extension of the constitutive model to account for crystal plasticity and anisotropic strength will be required for a high‐fidelity continuum description.