Research Projects

Extreme events—rare, high-magnitude deviations from average system behavior—occur across a wide range of spatiotemporally complex systems, including turbulent flows, thermal convection, and combustion. These phenomena, such as rogue ocean waves, abrupt flow reversals in geophysical systems, and flashbacks in hydrogen combustors, often develop rapidly and unpredictably, making them particularly difficult to control. Despite their rarity, the consequences of extreme events can be devastating, affecting safety, efficiency, and operational integrity in both natural and engineered systems. While recent studies have made progress in identifying extreme events using data-driven tools or dynamical systems theory, few have bridged the gap between predictive theory and real-time experimental control. This project aims to address this critical gap by developing a robust, experimentally validated framework for the real-time prediction and mitigation of extreme events. The scientific objectives of this research are fourfold: first, to identify precursors to extreme events using both first-principles-based and data-driven approaches; second, to uncover the dynamical origins of these events using tools from nonlinear dynamical systems theory; third, to develop and deploy real-time, closed-loop control strategies in laboratory settings; and fourth, to build a real-time Particle Image Velocimetry (PIV) system to enable fast flow state measurements and rapid decision-making. These objectives will be pursued across three carefully selected test cases that span both fundamental and applied domains: quasi-two-dimensional shallow flows (representative of oceanic turbulence), Rayleigh-Bénard convection (modeling atmospheric circulations), and hydrogen combustion in reheat burners (an industrially relevant scenario). At the core of this project is the hypothesis that extreme events arise when a high-dimensional complex system temporarily departs from a statistically steady regime by following unstable directions in phase space. This hypothesis will be tested through a combination of simulations and laboratory experiments. The team will employ modal decomposition techniques (such as proper orthogonal decomposition and co-kurtosis PCA), neural network-based methods (including autoencoders and reservoir computing), and stochastic modeling to identify precursory patterns and develop forecasting tools. By integrating these prediction models into laboratory setups and developing real-time control protocols—such as electromagnetic forcing in fluid flows or thermal actuation in convection cells—the project will evaluate whether early, low-amplitude interventions can effectively prevent or mitigate extreme events. The anticipated impact of this project is substantial. From a fundamental science perspective, it promises to advance our understanding of the mechanisms driving extreme events, validate theoretical models in experimental settings, and create a generalizable framework applicable across disciplines. From an application standpoint, the project’s outcomes could inform the design of safer combustion systems, contribute to early warning systems in climate modeling, and improve operational reliability in systems where extreme fluctuations are detrimental. Additionally, the development of a real-time, open-source PIV system within an academic environment offers significant economic and research benefits, particularly for the broader Indian fluid dynamics community. By combining state-of-the-art experimentation, theory, and machine learning, this project seeks to make a lasting contribution to the science and control of extreme events.