Thesis

FR EN

Inverse Problems in Atmospheric Dispersion - Application to Source Characterization in Complex Industrial Sites - J. SALLES LOUSTAU

The main objective of this thesis work is to develop an inverse modeling tool, capable of representing the complexity of transport and dispersion phenomena at a complex industrial site and inverting the characteristics of emissions and leaks, given different types of on-site measurements.
The objective of the thesis project is to develop an inverse modeling of atmospheric dispersion of pollutants, suitable for the study of complex industrial sites, based on adjoint formulations (Liu et Zhai, 2007, Müller et Stavrakou, 2005, Quélo et al., 2005) of the PYRAMID software models.
In addition, we propose to investigate machine learning approaches (Chen et al., 2020, Cao et al., 2019), with the objective of characterizing sources in real time in the event of an unexpected leak. While these approaches rely on an expensive training phase from accurate synthetic data, they then allow a very efficient application for real-time predictions; provided of course that the learning was sufficiently accurate.
Thesis funded by TOTAL ENERGIES - LMFA

Assimilation of mobile concentration measurements by micro-sensors for high resolution simulation of air quality in the environment urban - M. OTALORA

Objective of this thesis is to develop and validate a methodology for assimilation of fixed or mobile concentration measurement data by low-cost μ-sensors, for the simulation of air quality in urban areas.
Thesis CIFRE : Aria Technologies and Laboratoire de Mécanique des fluides et d'acoustique (LMFA)

Atmospheric dispersion modeling and data assimilation of pollution optical measurements : application to the exploitation of measurement campaigns on complex industrial sites - J.B MEJIA ESTRADA

Nowadays the control and surveillance of gas and particles emitted into the atmosphere is done more through optical systems, e.g. hyper-spectral cameras, lidars, satellites images, etc. Optical instruments allow the measurement of pollutant concentration in the plume. Moreover, the latter could be used to improve the modelled concentration field and, indirectly, to estimate the emissions. Nevertheless, the method used to identify and quantify the emission source affects the magnitude and the uncertainty of the predictions.
The use of optical measurements could contribute to:
  • A better knowledge of industrial sites.
  • Management of the future regulatory controls, made through remote sensing addressed to measure COV and methane
  • Improvement of physical representability of dispersion models employed for prevention of major risks
  • An approach validated for critical environmental events
  • Recognition of emission sources
    In order to make the most of the new experimental data, characterised by a high sampling frequency and a strong level of fluctuations, a robust direct simulation approach is required. For this reason, the first part of the current thesis is focused on the study of the atmospheric boundary layer and the improvement of its CFD simulations using RANS and LES models. Above all, introducing LES simulations and optical measures into operational applications requires a careful analysis of the available resources. The computational time, directly linked to computational power, and the amount of CFD data generated are critical. For this reason, it is very important to accurately choose a CFD code and direct toward a database management system. A suitable wind field also allows the dispersion models to be enhanced, both Eulerian and Lagrangian. Moreover the description of pollutant ejection characteristics impacts significantly on the near field. It implies the development of a specific module which manages this issue.
    The second part concerns the development of data assimilation and inverse methods using the optical measurements. The aim is to develop a solid technique to assimilate different kinds of optical measurements into the numerical model and vice-versa. In fact, the direct model could benefit from measurements through a proper spreading of local corrections along the domain. In the same way, the plume reconstruction made by two or more cameras could profit from model in order to obtain a more realistic 3D reconstruction. Finally, the inverse problem represents a big challenge, the use of classical and sophisticated optimisation algorithms would be employed to wholly characterise the emission source.
    Thesis CIFRE: Laboratoire Qualité de l’Air, TOTAL and Laboratoire de Mécanique des Fluides et d’Acoustique (LMFA)
  • Lagrangian simulation of the atmospheric dispersion of a rejection unsteady in the presence of obstacles. Application to accidental and malicious releases of pollutants in the environment urban - M. SLIMANI

    One of the major topics concerning atmospheric dispersion is the release of toxic substances (Nuclear (N), Radiological (R), Bacteriological (B) or Chemical (C)) for the safety of people and infrastructures. For several years, terrorist risks and malicious acts were add to the context of industrial risks. This thesis aims to answer 3 problems :
  • How to take into account the topographic complexity of urban, peri- urban or industrial buildings in a rapid operational modeling of atmos- pheric dispersion ?
  • How to describe the turbulent variability of the possible behaviors of an atmospheric plume in a situation of short discharge ?
  • How to characterize by inverse modeling the emissions of a rejection a priori unknown ?
    To answer the problems, the thesis project will develop a new generation of textsc Siranerisk by adopting the Lagrangian formalism allowing a quick response while validating the results through comparisons with databases.
    Thesis funded by CEA, TOTAL ENERGIES - Laboratoire de Mécanique des Fluides et d’Acoustique (LMFA)

    See all thesis in AIR - Atmosphere, Impact & Risk