Simplified Models for Fire-Atmosphere Interactions

An important factor in fire and plume spread is the complex and multiscale fire-atmosphere interaction. One way to describe these complex interactions is with simplified physics- and data-driven models. I will describe several models for pyrogenic potentials and fire-induced vorticity, and will demonstrate their effect on fire and plume spread. Even though these simplified models are computationally faster than a full physics-based model, they are still too slow if large numbers of simulations are needed. This happens, for example, when performing detailed statistical analyses or solving inverse problems. Therefore, to further accelerate these simplified physics-based models, they will be trained with neural network architectures which results in a significant speedup.