Job offer
- Discuss this with your agent
- Open in Claude
- Open in ChatGPT
Prepare application
- Draft this with your agent
- Open in Claude
- Open in ChatGPT
Role details
Tech stack
+3 more
Job description
The ALÉAS project (“Extreme turbulence: predicting the unpredictable”), led by LPENSL, SPEC (CEA Saclay) and LEGI (Grenoble), aims to detect the weak signals that precede extreme events and abrupt transitions in turbulent flows. Transitions between metastable states are rare events, occurring over times much longer than the characteristic time scales of turbulence: neither theory, nor simulation, nor learning from conventional databases can currently anticipate them. The project relies on a breakthrough metrology: neuromorphic imaging (event-based cameras), which records only the pixels whose intensity changes. LPENSL has developed a 3D Lagrangian tracking method based on eight event-based cameras. It was used to record continuously, for 24 h at 10 kHz, a turbulent von Kármán flow in a square vessel (Reλ 100), yielding 5.2 million trajectories and 5 billion acceleration samples. For the first time, this recording continuously resolves the bistability of the flow: 53 spontaneous switches between two large-scale states, with a median duration of 24 min. The postdoctoral researcher’s mission is to search for precursors of these transitions from the neuromorphic measurements. This will involve deploying machine-learning methods to (i) identify the state of the flow, (ii) detect and characterise the transitions and (iii) develop strategies to search for precursors. The possibility of generating transition data numerically, through simulations in a geometry close to the experiment, will also be explored.
The postdoctoral researcher will exploit the long recordings already acquired and take part in new measurement campaigns on the LPENSL von Kármán experiment. Main activities:
- State identification: build descriptors of the large-scale state from Lagrangian trajectories (possibly directly from the event streams) and develop supervised or unsupervised classification methods (dimensionality reduction, Gaussian mixtures, hidden Markov models, neural networks).
- Transition detection: accurately date the switches (change-point detection), characterise transition paths and the statistics of residence times.
- Precursor search: analyse statistics conditioned on transitions (intense acceleration events, local anisotropy, large-scale fluctuations), test early-warning indicators and develop predictive models, from autoregressive models to their nonlinear machine-learning generalisations, quantifying their prediction horizon.
- Experimental campaigns: contribute to new long-duration recordings (several days, various forcing conditions) and to their coupling with other measurements (torque, acoustics, pressure) that may carry precursor information.
- Numerical simulations: explore, with the LEGI and SPEC partners, the generation of transition data by simulation (large-eddy simulation, log-lattice models) in a geometry close to the experiment, and their comparison with measurements.
- Dissemination: publish results in international journals, present them at conferences and take part in the scientific life of the project.
The postdoctoral researcher will be based at the Laboratoire de Physique de l’ENS de Lyon (UMR 5672, CNRS / ENS de Lyon), under the supervision of Mickaël Bourgoin and Nicolas Plihon, within the team that developed 3D Lagrangian tracking with neuromorphic cameras. The laboratory hosts the square-vessel von Kármán experiment and eight event-based cameras, and has access to the computing resources of the PSMN mesocentre. The ALÉAS project brings together three partners with complementary expertise:
- LPENSL (Lyon): experiments, Lagrangian and neuromorphic metrology;
- SPEC (CEA Paris-Saclay): theory and von Kármán experiments, bifurcations and extreme events;
- LEGI (Grenoble): turbulence modelling and simulation, machine learning for simulation. The postdoctoral researcher will interact regularly with the PhD students and postdocs recruited at the other project sites and may make research stays at SPEC and LEGI.
Requirements
PhD or equivalent
Research Field Chemistry
Education Level PhD or equivalent
Languages FRENCH
Level Basic
Research Field Physics
Years of Research Experience 1 - 4
Research Field Chemistry » Computational chemistry
Years of Research Experience 1 - 4, * PhD in physics, fluid mechanics, nonlinear physics or data science applied to physics
- Strong experience in machine-learning methods (classification, dimensionality reduction, deep learning) and time-series analysis
- Proficiency in Python and scientific and machine-learning libraries (NumPy, scikit-learn, PyTorch or equivalent)
- Experience in processing large datasets (several terabytes) and in cluster computing
- Knowledge of turbulence, multistable dynamical systems or rare events appreciated
- Experience in Lagrangian particle tracking, high-speed imaging or numerical flow simulation (LES, DNS) would be an asset
- Interest in working at the interface between experiments, modelling and numerics
- Good writing skills and fluency in scientific English
- Autonomy, rigour and ability to work as part of a multi-partner project
Apply for this position
This job is hosted externally. Click below to view the full posting and apply.
Prepare application
- Draft this with your agent
- Open in Claude
- Open in ChatGPT
Good distractions
Loading talks and stories from around this role…