Required Education : Master 2 ou Diplôme d'ingénieur ou Equivalent
Start date : 1 October 2019
Mission duration : 3 ans
Deadline for applications : 15 June 2019
Salary : environ 2439 euros brut/mois
The objective of the PhD thesis is to design a methodology to give access to more accurate numerical simulations of plume dynamics, which could be used to assess risks and more particularly to identify the critical areas that are associated with high tracer concentration levels. For this purpose, the key idea is to design an emulator of three-dimensional large eddy simulations (LES) which accurately and efficiently, predicts quantities of interest and accounts for uncertainties. Pollutant dispersion simulations are subject to a large number of parametric uncertainties. The emulator will provide a framework to account for these parametric uncertainties. The PhD thesis follows two lines:
(1) build a reduced-order model (ROM) that is adapted to represent micro-scale atmospheric dynamics and build a model of the ROM error through a automatic deep learning strategy;
(2) build an efficient method to account for structural and parametric uncertainties through the construction of an emulator (also known as metamodel or surrogate) of the ROM and through the definition of quality metrics that are adapted to the pollutant dispersion problem.
The PhD thesis includes LES, uncertainty quantification and statistical learning. This will be done in collaboration between Cerfacs (Centre Européen de Recherche et Formation Avancée en Calcul Scientifique) et le LIMSI (Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur, CNRS/Université d'Orsay). The PhD advisors will be Mélanie Rochoux (Cerfacs) et Didier Lucor (CNRS). Thomas Jaravel (Cerfacs/équipe GLOBC) and Olivier Vermorel (Cerfacs/équipe CFD) will also be involved.
The PhD student will be employed by Cerfacs with a salary of about 2439 euros gross per month. The PhD thesis will start in Fall 2019 (October or November, this could be discussed). The PhD thesis will be mainly carried out at Cerfacs (Météo-France cmpus, Toulouse, France). Regular stays at LIMI (campus Paris-Saclay) are planned during the PhD. Regular visits of Didier Lucor at Cerfacs are also planned.
This multidisciplinary project is considered as innovative for the design and development of advanced methods for public and/or industrial safety, at the interface between multiple expertises present at Cerfacs and LIMSI. The PhD student will benefit from this environment. Furthermore, the PhD thesis has strong links with the PPM project funded by RTRA-STAE and piloted by Cerfacs. This project gathers research institutes such as CNRM (Météo-France), ONERA and CEREA (EDF/Ecole des Ponts ParisTech). The general objectives of the PPM project aims at providing methodologies to run ensemble simulations related to micro-scale applications at the scale of a “site” (urban area, industrial site). Research related to the study of turbulent natural convection with statistical learning approach in the context of heat transfer intensification is also ongoing at LIMSI.
The PhD candidate should have a master/engineering degree or equivalent, with a strong background in fluid mechanics and/or statistical learning, or any related field. Mathematical and theoretical knowledge would also be appreciated, as well as some experience with Fortran/Python codes. Oral and writing skill in English is mandatory, French would be a plus. Furthermore, the PhD candidate should be a good team member and take initiatives to carry out this PhD thesis.
If you are interested, please send a cover letter, a CV and two recommendation letters to the advisors
- Mélanie Rochoux, email@example.com
- Didier Lucor, firstname.lastname@example.org (https://perso.limsi.fr/lucor/index.html)