Fast identification of accidental pollution based on an innovative geometrical method with 3D-CFD modelling - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Fast identification of accidental pollution based on an innovative geometrical method with 3D-CFD modelling

Damien Joseph
  • Fonction : Auteur
Malo Le Guellec
  • Fonction : Auteur
Claude Souprayen
  • Fonction : Auteur
Jean-Pierre Issartel

Résumé

This paper focuses on the implementation of a new method for the localisation and identification of a source of biological pollutants as part of the continuous monitoring of a site with hazardous material or processes. The origins of this method is the renormalisation theory developed for the French military (DGA) to localise nuclear, radiologic or chemical emissions. However, it can be adapted for industrial use for major chemical platform. Currently, most methods for inverse modelling and source identification are based on a Bayesian approach. However, these methods have both theoretical and conceptual limitations, and, uncertainties about their performances, which make them hardly adequate to the problem of detection and identification of NBC releases. The renormalization theory allows, through the precise knowledge of wind fields, to identify a source of pollution from atmospheric tracer, provided by a network of sensors to various instants over the area of interest. Because a sensor network does not perceive all the points of the environment fairly in Euclidean geometry, a new geometry is determined from the correlation of the adjoint functions from the various sensors providing a mapping of sensitivity of the sensor. The quality of the reconstruction depends on the good modeled representation of the physical dispersion proposed. Its operational implementation in a robust monitoring system coupled with precise 3D CFD modelling is essential. In order to assess the impact of the concentrations fluctuations expected at real sensors on the performance of the source reconstruction by renormalization theory in terms of location and mass flow rate estimation, LES simulations of a grid turbulence case have been performed and analyzed to assess the expected error due to a RANS dispersion model which is used for the adjoint functions. The results analysis in terms of averaged concentration, concentration integration and standard deviation shows that with an appropriate sampling system, the renormalization approach based on a RANS model give a good reconstruction of the source even with noisy data related to complex geometry and related turbulence.
Fichier non déposé

Dates et versions

hal-01178989 , version 1 (21-07-2015)

Identifiants

  • HAL Id : hal-01178989 , version 1

Citer

Damien Joseph, Malo Le Guellec, Claude Souprayen, Jean-Pierre Issartel, Pierre Ngae. Fast identification of accidental pollution based on an innovative geometrical method with 3D-CFD modelling. 16th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes (HARMO 2014), Sep 2014, Varna, Bulgaria. pp.566--571. ⟨hal-01178989⟩

Collections

UNIV-EVRY LMEE
83 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More