StaBility-integratEd machiNe lEarning For actIve flow conTrol – BENEFIT

This research proposal seeks to develop an innovative framework for designing active flow control strategies by synergistically integrating stability analysis, grounded in the linearized Navier-Stokes equations, with machine learning algorithms exploiting empirical flow data. The primary objective is to harness inherent flow instabilities to optimize specific quantities such as friction, heat transfer, or other relevant parameters while minimizing the energy input required for effective control. The framework will leverage the strengths of both model-driven and data-driven methods, combining the extensive expertise of the aerodynamics community in stability analysis with the emerging capabilities of machine learning.

The research program will focus on flows over concave walls, where Görtler vortices naturally occur, providing an ideal testing ground for the proposed control strategies that aim to manipulate flow instabilities for optimal flow control. This interdisciplinary approach aims to advance the understanding of near-wall turbulence and flow control while addressing the pressing need for efficient, cost-effective energy transfer in engineering applications. The project is organized around four research areas : coupling stability analysis with machine learning, reduced-order modeling, design and implementation of perturbation strategies, and application of control laws.

Through the synergistic combination of advanced numerical simulations, experiments, and machine learning, the project introduces a new paradigm for extracting information on flow dynamics, building robust models, and designing optimal control strategies. The consortium, composed of researchers from IUSTI, Pprime, and DynFluid laboratories, brings together complementary expertise in experiments, numerical simulations, stability analysis, and machine learning algorithms to deliver disruptive physics-based data science tools for control.