Atmospheric dispersion of pollutants on an industrial facility

Context

To anticipate and prevent technological and industrial risks, it is necessary to assess the consequences of the atmospheric dispersion of pollutants according to the characteristics of the leak, atmospheric conditions and site topology, and to contain any leak as quickly as possible.

However, on the scale of an industrial site, the specifications are sometimes complex:

• By the scale of the area to be modeled, from a few hundred meters to several kilometers,
• By the multiplicity of substances that can be released, at several points on the site,
• By the detection time scales, which, for the protection of neighboring areas, are generally of the order of a minute,
• By the need to double the network, i.e. each cloud must be detected by at least 2 sensors,
• By the cost of purchasing and maintaining sensors, as well as the increased risk of failure when the number of sensors is higher.

Objective

In the case studied here, the customer is initially planning to install 1,200 sensors in order to meet the technical specifications point by point. The aim of the study carried out here is multiple:

• To accurately simulate the most critical pollutant clouds, based on site topography, leakage and meteorological conditions,
• Based on these critical scenarios, to determine the optimum sensor positions in order to meet specification criteria while minimizing the number of sensors.

Simulation and results

Benefits of CFD simulation for atmospheric dispersion studies

Atmospheric dispersion studies generally involve 3 main steps:

• Calculation of the source term, which consists in determining the characteristics and conditions of the release: flow rate, velocity, temperature, etc.
• Calculation of the atmospheric dispersion of the pollutant, which involves determining the movement and concentration of the cloud as a function of the source term, the site and meteorological conditions.
• And finally, the calculation of the effects on people, which consists in analyzing the risks according to the nature of the pollutants, their concentration and exposure times.

To calculate the source term, abacuses or analytical methods generally exist (such as the INERIS Omega 19 guide). However, when the release configuration goes beyond the predefined framework of analytical methods, numerical simulation can be used to model the characteristic properties of the toxic or flammable leak (flow rates, pressure, liquid fraction, etc.), and then use these results in the next phase.

Indeed, it is mainly when it comes to modeling the atmospheric dispersion of pollutant clouds that CFD simulation offers major advantages. Compared with the 1D Gaussian models typically used for this type of study, 3D CFD tools make it possible to take into account the influence of wind direction, site topology, the presence of buildings and other obstacles to dispersion… They are particularly well-suited to modeling the atmospheric dispersion of pollutants in the near field (< a few kilometers) and/or in the presence of obstacles, buildings or uneven terrain.

Validation and optimization

In this case, the numerical simulation was used to simulate all the critical scenarios identified.

These scenarios were used to generate a substantial volume of data (x,y,z,t) containing the position of the cells in the mesh for which the concentration is above the detection threshold, at each instant, for each scenario.

Mathematical optimization methods were then implemented, enabling the number of sensors to be reduced by around 90%, while still perfectly meeting the initial specifications.