Nom du directeur de thèse : Pierre Boivin
E-mail : pierre.boivin@univ-amu.fr
Co-encadrant et contact : Grégoire Varillon
E-mail : gregoire.varillon@univ-amu.fr
Financement : demandé / Type de financement : école doctorale
Context:
Coherent structures in turbulent flows appears as an organized behavior dynamical structures among the chaotic flow motion. The role of coherent structures in many applications is para-mount as they are linked to unstable flow behavior (e.g., vortex shedding), acoustic interactions, low frequency noise or fluid-structure interactions. Their development can be detrimental -- e.g., structural vibrations – or beneficial, as a small initial fluctuation can be largely amplified by flow instabilities.
The Linearized mean-fields method is an efficient computational methods for identifying them: the most energetic structures appear as the dominant eigenmodes, or global modes, of the Navier-Stokes Equations linearized (LNSE) around a mean flow. In the frequency domain, the resolvent analysis identifies the dominant flow structures and amplification mechanisms at a given frequency. These eigenmodes are often computed with the adjoint eigenmodes that contain gradient information used for sensitivity analysis or optimisation. The resolvent modes also provide forcing locations for optimal control, and yield a factorization of the evolution operator for coherent structures into a low-order model.
Scientific bottleneck and objective of the thesis:
However, applying these methods to large geometries is out of reach due to memory and computational time limitations. This thesis will leverage direct-adjoint looping and matrix-free strategies together with the efficiency of lattice Boltzmann methods (LBM) for achieving Linear-ized mean-field methods on large and complex geometries. The outcome of the thesis are:
- A well-posed and robust methodology for performing resolvent analysis and global mode analysis with LBM,
- An efficient workflow for computing these modes on large configurations,
- A flow control method via optimal forcing for large configurations with LBM.
CANDIDATE
You earn a M.Sc. degree in fluid dynamics, or with a significant training in fluid mechanics, specifically with advanced knowledge on turbulent flows and their modeling.
Technical skills:
- Must-have: an experience in Computational Fluid Dynamics (CFD) and good knowledge in numerical methods applied to CFD,
- Must-have: scientific computing, experience in code development and good coding prac-tice,
- Nice-to-have: experience or knowledge in linear stability analysis or optimization algorithm, experience with C++.
Soft skills: proficiency in English, synthetic writing, eager to communicate on your topic orally and written, motivation and independence for problem solving.
Your preferences: numerical method development, code development, physics-based modeling.
