Research in Fluid Mechanics

A significant part of my research explores active flow control techniques, primarily spanwise wall oscillations, to reduce turbulent skin friction drag. Detailed DNS studies have allowed me to elucidate the underlying mechanisms, including the disruption of near-wall streaks by the unsteady Stokes layer, and the crucial roles of velocity skewness and hysteresis in the drag reduction cycle.

I've identified that the mechanism behind drag reduction is not streak destruction (as previously thought) but rather prevention of streak regeneration. The spanwise Stokes layer induced by wall motion causes tilting of wall-normal vorticity tubes, which inhibits the streak formation process. This understanding is crucial for optimising control strategies.

I investigate why the effectiveness of such control strategies diminishes at higher Reynolds numbers, linking this decline to the increasing influence of large-scale outer structures (modulation). Current work extends this to heat transfer, exploring how wall oscillations can preferentially enhance heat transfer over drag, potentially breaking the classical Reynolds analogy (linked to ANR SOLAIRE project).

My recent findings show that with carefully selected actuation parameters (T+ = 500, W+ = 30), heat transfer can be enhanced by 15% while drag increases by only 7.7%, challenging conventional understanding that heat transfer and momentum transfer behaviours are tightly coupled. This discovery opens new possibilities for thermal management applications.

Key Points:

• Analysis of drag reduction mechanisms via spanwise wall oscillations.
• Identification of the roles of skewness, hysteresis, and Stokes strain.
• Quantification of Reynolds number impact on control efficiency due to large-scale modulation.
• Investigation of heat transfer enhancement and breaking the Reynolds analogy using wall oscillations.
• Development of reinforcement learning approaches for optimising control parameters.

Understanding how large-scale outer structures influence the near-wall region (footprinting and modulation) is critical for predicting high-Reynolds-number flows. My work scrutinises these interactions, revealing that the "modulation" effect is not a direct scale interaction but rather an indirect consequence of large scales altering local shear production rates.

I have shown that the response of small scales is asymmetric: positive large-scale fluctuations (sweeps) cause stronger amplification than the attenuation caused by negative fluctuations (ejections), partly due to "splatting" effects near the wall. This challenges linear predictive models like the PIO model proposed by Marusic and colleagues.

By developing novel statistical approaches based on multivariate joint PDFs and bi-dimensional Empirical Mode Decomposition (BEMD), I've demonstrated that the modulation of near-wall turbulence involves not just amplitude changes but also asymmetric redistribution of energy (splatting). This has led to improvements in predictive models for high-Reynolds-number flows.

I also investigate the validity of the Quasi-Steady Hypothesis (QSH), finding that near-wall statistics often scale well with the local large-scale friction velocity, particularly for amplitude modulation, but with limitations for wavelength modulation and at positions further from the wall.

Key Points:

• Analysis of amplitude and length-scale modulation mechanisms.
• Demonstration of asymmetric modulation and the role of "splatting".
• Quantification of scale contributions to skin friction using FIK and RD identities.
• Testing the validity and limitations of the Quasi-Steady Hypothesis (QSH).
• Development of improved predictive models for high-Reynolds-number flows.

The increasing volume and complexity of simulation and experimental data necessitate advanced analysis tools. I actively develop and apply data-driven methods, particularly machine learning, to extract insights, build reduced-order models (ROMs), and discover control laws for turbulent flows.

Key techniques include Bi-dimensional Empirical Mode Decomposition (BEMD) and Auto-Encoders (AE) for scale separation and feature extraction, especially in high-Re flows where traditional methods struggle. I've demonstrated how AEs can be used to identify large-scale structures in DNS data at Reτ ≈ 5200, where conventional methods become computationally prohibitive.

I also employ symbolic regression (e.g., Gene Expression Programming - GEP) to derive interpretable physical models from data, and explore clustering and Deep Reinforcement Learning (DRL) for ROM development and control optimisation (linked to ANR INFERENCE and MUFDD projects).

My approach to ROM construction involves a 5-step framework: (1) dimensionality reduction with auto-encoders, (2) clustering to identify key flow states, (3) development of Markov chain models for state transitions, (4) reconstruction of high-fidelity data, and (5) extraction of physical insights. This framework has been successfully applied to cylinder flows and is being extended to wall turbulence.

Key Points:

• Development of Auto-Encoder methodology for scale separation in large DNS datasets.
• Application of BEMD for analysing scale interactions.
• Use of multivariate joint PDFs for conditional statistical analysis.
• Exploration of GEP, Clustering, Markov Chains, and DRL for modelling and control.
• Development of a comprehensive ROM framework combining physics understanding with data-driven approaches.

A significant portion of my current research focuses on understanding and optimising heat transfer in turbulent flows. This work bridges fundamental fluid mechanics with practical thermal management applications, addressing critical challenges in energy systems and sustainable technologies.

My research on heat transfer spans both passive and active control techniques. The passive approaches utilise surface features and natural flow instabilities, while active methods employ techniques like spanwise wall oscillations. The latter has been extensively studied for drag reduction, but my work shows it also has untapped potential for heat transfer enhancement.

A key finding from my research is the possibility of "breaking" the Reynolds analogy—the traditional understanding that heat and momentum transport are tightly coupled in turbulent flows. Through careful selection of control parameters, I've demonstrated that it's possible to enhance heat transfer more substantially than the associated drag increase, which has significant implications for thermal management in various engineering applications.

In my ANR SOLAIRE project, I'm applying these insights to solar receivers, where efficient heat transfer is crucial for overall system performance. By manipulating wall-bounded turbulence, we aim to achieve substantial improvements in receiver efficiency without prohibitive pressure drop penalties.

My current work in this area also explores how the large-scale structures in high-Reynolds number flows influence thermal transport, extending predictive models to include both momentum and heat transfer. This research is driven by both fundamental scientific interest and the urgent practical need for more efficient thermal management solutions as we transition to sustainable energy systems.

Key Points:

• Investigation of the relationship between momentum and heat transfer in wall-bounded flows.
• Discovery of active control parameters that preferentially enhance heat transfer over drag.
• Analysis of the underlying physical mechanisms that allow breaking the Reynolds analogy.
• Application to solar receiver design in the ANR SOLAIRE project.
• Development of machine learning optimisation approaches for thermal management.

An innovative aspect of my research involves exploiting natural flow instabilities, particularly those that develop on concave surfaces, to enhance heat transfer with minimal energy input. This approach leverages Görtler instabilities that naturally form coherent vortical structures when a fluid flows over a concave surface.

Unlike traditional passive vortex generators (such as fins or ribs) that increase pressure drop significantly, my approach uses minimal perturbations to trigger and control these natural instabilities. By carefully tuning these perturbations using plasma actuators, we can generate longitudinal vortices that significantly enhance mixing between the wall and free-stream fluid.

This research has important applications in heat exchanger design, particularly for aerospace applications where both thermal performance and weight/pressure drop considerations are critical. In collaboration with industrial partners like Safran, we're exploring how these concepts can improve thermal management in aircraft engines.

The control strategy involves more than just triggering instabilities—it's about precisely manipulating vortex characteristics (size, spacing, strength) to achieve optimal heat transfer enhancement while minimising the associated pressure penalty. My research employs both high-fidelity simulations and experiments to develop and validate these approaches.

Key Points:

• Exploitation of Görtler instabilities on concave surfaces for efficient vortex generation.
• Development of plasma actuation techniques for precise control of vortical structures.
• Optimisation of vortex characteristics for maximum thermal enhancement with minimal pressure loss.
• Application to aerospace heat exchangers in collaboration with industrial partners.
• Combination of high-fidelity simulations and experimental validation.