A Preliminary Look at the Effects of Model Complexity on the Outcomes of Fire and Smoke Simulations

In numerical modeling, the computational resources and execution time required for a simulation generally depend on the size of the model domain, the intricacy of the underlying physics, the spatial and temporal resolution of the modeled phenomena, and the desired time interval of simulation. Broadly speaking, computational expense increases with model complexity. However, depending on the phenomenon and spatiotemporal scales of interest, the physical fidelity that might be gained by adding more physics or increasing resolution may not have a significant impact on the utility of the simulation results. Modeling systems that can incorporate different and various sub-models, and resolve the domain at multiple scales, provide a platform to address the question, “how much is enough?” when simulating physical phenomena.

An example of such a system is NIST’s Fire Dynamics Simulator (FDS), which can simulate combustion phenomena over scales ranging from millimeters to kilometers. We’ll present FDS simulations indicating that a relatively simple and fast empirical sub-model of fire propagation, if coupled with the surrounding flow, can produce results consistent with a more realistic, though computationally demanding, physics based (PB) sub-model. In addition, an analysis of FDS plume simulations will be presented to clarify the role of source resolution and spatial distribution on smoke dispersion. Results will also be compared to empirical plume formulas, helping us to determine “how much is enough?” when modeling dispersion. Other issues related to model uses and limitations will be discussed as time allows.