Modeling of Toxic Industrial Chemical releases in realistic environments.

Toxic Industrial Chemicals (TIC) are often quoted as the most hazardous materials and constitute a severe threat for people living in proximity of industries and storage areas where such materials are handled. For example, three years ago, a large release of chlorine gas in the Chinese province of Jiangsu caused the death of 27 people. Releases of TIC into the atmosphere lead to the formation of a dense gas cloud containing a mixture of liquid droplets and gas. For realistic risk analysis purposes, disper sion models must have an accurate prediction of the conditions at the rupture, the complex interactions between the two-phase cloud, the atmospheric turbulence and eventual obstacles around the release. At present, the influence of obstacles and ground o n TIC releases is neglected by the modelers. A model that handles the complex physic of TIC releases in realistic environment will be developed and implemented in the 3-D Computational Fluid Dynamics code FLACS which can compute the flow and dispersion of pollutants around obstacles. FLACS is a world leading tool for gas explosions modeling and is used by more than 50 companies worldwide. Two different approaches might be employed to model the behavior of a two-phase cloud. A Lagrangian approach can be us ed to track each liquid droplet or a collection of droplets with equal properties and model the droplets impact on the obstacles. On the other hand, an Eulerian approach can be developed in which the liquid droplets are defined as a third continuum in add ition to the gas and air components. The strategy chosen in this research project is to first improve and evaluate the performance of the Eulerian approach, which is already implemented in a R&D version of FLACS. If the Eulerian approach shows to be insuf ficient in order to capture the complex physics, a droplet tracking model will be implemented. Finally, some effort will be given to evaluation and improvement of the atmospheric boundary layer modelling