(115c) Stochastic Programming and Atmospheric Dispersion Modeling Dispersion Modeling for Optimal Placement of Gas Detectors

Tragic accidents like the Buncefield fire are harsh reminders of the need for proper detection and mitigation in case of the emergence of incidents. Installing gas detectors at appropriate locations are one of the indispensable conditions for the implementation of emergency response plans. With strong global interest, optimal placement of gas detectors based on Computational Fluid Dynamics (CFD) simulations is developed. In this research, we introduce investigations aimed at optimizing the placement of gas detectors (fixed and mobile) and using tools for calculating atmospheric dispersions. There are significant uncertainty to the various factors to be considered: process conditions, the impact of surrounding geometries on dispersion, weather conditions, physical limitation to land use, etc. The uncertainty associated with different leak scenarios is captured through process specific CFD simulations using a package for rigorous modelling and simulation of gas dispersion. Using simulation results from the dispersion modeling, a multi-scenario, mixed-integer linear programming formulation is developed to generate optimal placement of gas detectors by minimizing the expected number of detection over all scenarios; this verifies our studies can be effectively applicable on any processes. The result of this study can help to place mobile sensors as well as fixed point-type or area sensors for better tracking of gas releases based on real-time monitoring of concentration data. Adoption of IoT in relation to cyber-physical systems is also discussed.

Key words: Chemical process safety, Hazardous material release, Source tracking, Accident response