Domain Decomposition for a Hybrid State Estimation of a Plume Field with a Moving Sensor

Le résumé fourni par la source

The estimation in real time of the plume concentration profile that results from a stationary or moving gas source releasing in the atmosphere is important for both national security and environmental protection. To estimate in real time the concentration profile and track the gas source, a parallelized heterogeneous non-overlapping domain decomposition explicit (HT-NODDE) method is developed and implemented in 3D. In the subdomain where the mobile sensor resides, a Luenberger estimator model is solved while in the other subdomains a naive estimator model is solved. The transmission conditions on the interfaces, serving as a consensus protocol between the Luenberger and naive observers, ensures that the naive observer receives process information. A structural grid with uniform cell length in each direction is used to discretize the entire domain. To maximize the efficiency of parallel computing, we keep the same number of cells in each subdomain. The advection-diffusion estimation equation is discretized in space by using a Finite Volume Method (FVM) with Total Variation Diminishing (TVD) scheme and the resulting semi-discrete equation is integrated with respect to time by using a 4th order Runge-Kutta method. Continuity and conservation transmission conditions are ensured on the interfaces between adjacent subdomains. A Lyapunov-based guidance law and a modified guidance law are used to reposition the mobile sensor for optimal information. The parallel algorithm is implemented in 3D for instantaneous and continuous source release cases.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.