H​/F chercheur "Stratégies multi-fidélité "goal-oriented" pour la prédiction des écoulements turbulents

Job Title

Developing a methodology to automatically select the most suitable simulation strategy for a given prediction objective, optimizing the trade-off between accuracy and computational cost.

The candidate's mission will be to develop a methodology to automatically select the most suitable simulation strategy for a given prediction objective, optimizing the trade-off between accuracy and computational cost. The work will be based on multi-fidelity modeling approaches combining simulations of different fidelity levels, ranging from RANS calculations to high-fidelity LES simulations, possibly associated with different levels of geometric or physical simplification.

Particular attention will be paid to taking into account uncertainties and defining criteria to quantify the contribution of each simulation to the improvement of the prediction. The developments will also aim to extend current approaches, often limited to integrated quantities, to the reconstruction of complete statistical fields. The final objective is to construct a "goal-oriented" strategy capable of determining, for a given quantity of interest (performance, stresses, losses, fluctuations, etc.),

which simulations should be performed and at what level of fidelity in order to achieve the desired accuracy at the minimal cost.

Skills expected:
Solid knowledge in numerical fluid mechanics. Knowledge of RANS and/or LES methods. Interest in quantifying uncertainties and multi-fidelity approaches. Ability to analyze and exploit large simulation databases. Aptitude for collaborative work in a multidisciplinary environment. Desired experience:
Experience in numerical simulation of flows. Experience in modeling, quantification of uncertainties or statistical methods. Experience in data science or statistical learning is an asset. Experience in high-performance computing is appreciated. Desired degree:
Doctorate in fluid mechanics, physics, applied mathematics, scientific computing or related field. Application dossier: A detailed curriculum vitae and a cover letter.

The Laboratory of Geophysical and Industrial Flows (LEGI) is a Mixed Research Unit (UMR 5519) of the National Center for Scientific Research (CNRS), the National Polytechnic Institute of Grenoble (Grenoble INP) and the University of Grenoble-Alpes (UGA). LEGI conducts a wide variety of activities with a common core of skills: research in fluid mechanics and transfers. Scientific context:
The numerical simulation of turbulent flows relies on a permanent trade-off between accuracy and computational cost. Depending on the objectives pursued, different modeling levels can be mobilized: RANS simulations, hybrid RANS/LES approaches, LES, as well as various geometric or physical simplifications. Each of these approaches has its own level of fidelity, numerical cost, and domain of validity. In many applications, several simulation strategies coexist to answer the same scientific or industrial question.

For example, in rotor/stator configurations, it is possible to resort to RANS simulations with a mixing plane, hybrid RANS/LES approaches, LES on simplified geometry, or LES on full geometry. The choice of the most relevant strategy is often based on the user's experience rather than on a quantitative methodology. The MOST team of the LEGI laboratory has been developing advanced simulation methods in YALES2 for several years, covering a wide range from RANS approaches to high-fidelity LES simulations with mesh adaptation.

In parallel, recent work has enabled the development of multi-fidelity modeling strategies combining numerous but inexpensive simulations with more accurate but more expensive simulations. The recruited person will be assigned to the MOST team. The research activities of the MOST team (Modeling and Simulation of Turbulence) concern the numerical prediction of turbulent and multiphase flows to both increase our knowledge of the fundamental properties of flows and to optimize industrial systems.

The team aims to develop all the scientific domains necessary for the numerical study of turbulent and multiphase flows:…