Frank14fandCursor
0ade812864
Track the research dossiers, section freezes, supporting manuscript materials, and round-aware agent controls so future drafting decisions can be reviewed across both repository mirrors. Co-authored-by: Cursor <cursoragent@cursor.com>
12 lines
3.5 KiB
TeX
12 lines
3.5 KiB
TeX
\abstract{
|
|
%\addtocontents{toc}{\vspace{1em}} % Add a gap in the Contents, for aesthetics
|
|
|
|
The concepts of cloaking and illusion, originating from optics and electromagnetics, have inspired methods to manipulate wave-based phenomena across various physical fields. While significant progress has been made in linear systems, extending these ideas to the highly non-linear domain of fluid dynamics presents a noticeable challenge. This research addresses this gap by investigating a novel paradigm: using Active Flow Control (AFC) to achieve hydrodynamic cloaking and illusion. This represents a significant departure from the conventional use of AFC, as its application has traditionally been confined to performance-oriented goals like drag reduction or lift modulation. Hydrodynamic cloaking aims to render an object hydrodynamically "invisible" by actively restoring its downstream flow to the undisturbed background flow. Hydrodynamic illusion further extends this concept, sculpting a flow into the district wake of an imaginary target object.
|
|
|
|
To explore these novel AFC concepts, we employ a fluidic pinball as the actuator, comprising three independently rotating circular cylinders arranged in an equilateral triangle. These cylinders serve as actuators whose rotational velocities are dynamically optimized to manipulate the surrounding flow. The system dynamics is simulated through high-fidelity numerical simulations at Reynolds numbers around 100, encompassing a diverse range of challenging inflow scenarios: steady uniform flow, periodic Kármán vortex streets, and no-periodic isolated vortices. A data-driven, model-free Deep Reinforcement Learning (DRL) framework is adopted to discover the complex control policies required, utilizing feedback from downstream velocity sensors that emulate the perception capabilities of marine organisms.
|
|
|
|
The numerical results demonstrate exceptional performance and validate the proposed framework. The DRL-controlled system successfully achieves robust hydrodynamic cloaking, restoring downstream flow characteristics to over 90\% similarity with the undisturbed background flow, as quantified by Dynamic Time Warping (DTW) analysis of sensor signals. This was achieved under both steady and unsteady inflow conditions. Furthermore, the fluid pinball system demonstrates a remarkable capacity for hydrodynamic illusion. In a steady flow, the pinball actively generated wake patterns characteristic of imaginary cylinders with different diameters, achieving similarity scores of up to 97.5\% and precisely replicating key hydrodynamic identifiers such as the Strouhal number.
|
|
|
|
To ground these numerical findings in physical reality, an experimental platform featuring a towing system in a water tunnel and custom-designed instrumentation has been designed and constructed. Preliminary tests have been conducted to validate the practical feasibility of the approach, while also highlighting the significant engineering challenges associated with high-sensitivity, low-noise force and flow sensing in this demanding regime. Future work will focus on refining this experimental validation, alongside pursuing a deeper mechanistic analysis of the learned control policies with the goal of extracting interpretable physical models. This research lays the foundation for a new class of intelligent fluid systems, with profound implications for minimizing the hydrodynamic traces of underwater vehicles, developing advanced bio-inspired stealth technologies, and contributing to the preservation of sensitive aquatic ecosystems.
|
|
|
|
} % abstract end |