KSEF RDE: A Cyberphysical Approach to Wind Field Estimation with Application to Pollutant-dispersion

Heat, mass, and momentum transport between the Earth's surface and the atmosphere is closely linked to turbulence that evolves in the atmospheric boundary layer. Prediction of this turbulent behavior is crucial for many applications, including predicting the dispersion of dangerous chemicals or contaminants in the event of an unplanned release. However, accurate simulation of the atmospheric boundary layer is made problematic due to terrain topology, vegetation, interaction between a multitude of turbulent scales, surface roughness, and non-stationary conditions. We propose a measurement-based adaptive flow-field estimation (AFFE) approach to address the challenge of accurate prediction of the turbulent flow field, and hence pollutant dispersion. Specifically, we propose to couple a deployable flow-velocity-measurement system with a computational flow-field model. As envisioned, AFFE includes a physical system, which consists of autonomous unmanned aerial vehicles instrumented with flow sensors, and a cyber system, which consists of an adaptive computational flow-field model. Together, the physical system and cyber system are capable of producing accurate synthetic flow fields, which can be used to predict pollutant dispersion. The enabling technology is the surrogate performance model adaptation technique, which was originally developed to control uncertain dynamic systems. In this project, we adopt the surrogate performance method but use it in the context of turbulence model adaptation. Our model adaptation algorithm uses physical flow-field measurements to continually update the turbulence model contained within a numerical simulation. This method aims to produce a realtime synthetic flow field, which can be coupled with a pollutant dispersion model to provide accurate estimates of pollutant location and concentration. Simple turbulence models coupled with model adaptation allow for real-time flowfield estimation and dispersion prediction. The objective of the proposed effort is to demonstrate this approach at the laboratory scale and obtain the preliminary results required to seek federal funding.