MEchanics GAthering –MEGA- Seminar: Dynamical reduced-order model applied to Navier-Stokes for transitions in multi-stable flows

Abstract: The wake behind a car-like body can be multi-stable: it leans to one side, stays there, and at a random moment jumps to the mirror state. Such transitions are rare, and their waiting time grows exponentially with the inverse noise intensity, so computing their statistics by direct numerical simulations of the stochastically forced Navier–Stokes equations is unaffordable for most flows. This talk presents a stochastic reduced-order model to study them, in which every realisation of the flow shares one mean field and a handful of time-evolving modes, and carries only a vector of coefficients of its own.  Tested on the planar sudden expansion, an ensemble of thousands of realisations with sixteen modes reproduces the transition statistics of the direct simulations at a fraction of their cost, and makes the switching visible: a cloud of points moving between two basins.

 

Biography: After a BSc in Physics at King's College London, I graduated from the University of Oxford with an MSc in Mathematical Modelling and Scientific Computing. I then spent a gap year writing a finite-element method solver in Python. In September 2025 I joined LFMI as a PhD student under the supervision of Edouard Boujo. Since then, I have implemented a stochastic reduced-order model based on dynamically orthogonal equations within the spectral-element method code Nektar++, to compute noise-induced transitions in multi-stable flows at a fraction of direct numerical simulation cost.