<p dir="ltr">Energetic materials contain microstructural defects like cracks, voids, grain boundaries, and interfaces which act as nucleation sites for ignition and detonation when shocked. Finite element (FE) models are currently unable to capture explicit microstructure with voids, cracks, and randomly oriented grains with representative mechanics, thermal conduction, and reactivity that exhibit the full shock to detonation transition (SDT). Modern computational efforts seek to accurately model material response while also balancing efficiency and speed. Work presented in this thesis will highlight all of these microstructural features, investigate mechanical and thermal response of each microstructure, connect these results to what is observed in other experimental and computational work, and bring computational modeling even closer to an efficient model that contains all processes necessary to replicate SDT.</p><p dir="ltr">In energetic materials (EM), voids are irregular in shape, but most computational work has focused on circular void collapse behavior. However, geometries that contain irregularities or corners are more likely to act as initiation sites due to stress concentrations. Validation and calibration of void simulations with experimental lengthscales and loading conditions is still limited. Plus, pore collapse modeling efforts at low impact velocities do not model fracture, and it is known that cracks cause more extreme temperatures than pores.</p><p dir="ltr">Other microstructure characteristics like cracks and grains have sub-micrometer length scale, and influence the mechanical and thermal response of materials under extreme conditions. However, approximations and coarse-graining must be applied to continuum FE simulations to fit length and timescales required to capture phenomena such as detonations that occur at a millimeter scale. With the use of machine learning (ML), numerical models can be trained on results of small-scale microstructure simulations and applied to larger length and time-scale simulations. The ML model follows Microstructure-Informed Shock-induced Temperature net (MISTnet) model and is trained upon stress, strain, temperature, pressure, and slip data and includes crystal plasticity, fracture, friction, an equation of state, and heat conduction. The ML model is able to predict temperature fields behind the shock, concentrations at grain boundaries, and the influence of grain orientation.</p><p dir="ltr">Accurate temperature values are extremely important to modeling EM because thermal hot spots (HS) are the main cause of ignition. Critical HS cause the chemical reactions which transition the shock front into a detonation, but many continuum models do not include chemistry in their framework. A 1-step Arrhenius reaction model is added to FE mechanics model to investigate the relationship HS have on the run to detonation (RTD).</p>