Flow mediated interactions between two cylinders at finite Re numbers

Abstract

We present simulations of two interacting moving cylinders immersed in a two-dimensional incompressible, viscous flow. Simulations are performed by coupling a wavelet-adapted, remeshed vortex method with the Brinkman penalization and projection approach. This method is validated on benchmark problems and applied to simulations of a master-slave pair of cylinders. The master cylinder's motion is imposed and the slave cylinder is let free to respond to the flow. We study the relative role of viscous and inertia effects in the cylinders interactions and identify related sharp transitions in the response of the slave. The observed differences in the behavior of cylinders with respect to corresponding potential flow simulations are discussed. In addition, it is observed that in certain situations the finite size of the slave cylinders enhances the transport so that the cylinders are advected more effectively than passive tracers placed, respectively, at the same starting position.

MRAG-I2D: Multi-resolution adapted grids for remeshed vortex methods on multicore architectures

“…This allows users to formulate simulations without the need of explicitly interacting with a highly sophisticated spatiotemporally-adaptive machinery. The present software have been used for the simulations of bluff body flows [23], flow-mediated interactions between two cylinders [24], for the analysis of optimal shapes in anguilliform swimming at intermediated Reynolds numbers [25], and the investigation of reinforcement learning within the context of self-propelled bodies [26].…”

Self-propulsion of a counter-rotating cylinder pair in a viscous fluid

“…The method and code have been extensively used and validated for similar problems. 8,[16][17][18][19] Throughout this work, we set the effective resolution of our unit square domain to E R = 32 768 2 , and the cylinder radius equal to R = 0.0025, so that there are over 80 grid points across the radius at the maximum level of grid refinement. The refinement and compression thresholds are r tol = 10 −3 , c tol = 10 −4 , and the cylinder is represented on the computational grid through a discrete Heaviside function.…”

“…[4][5][6] Alternative forms of aquatic locomotion have been presented for a flapping body moving laterally due to vortical instabilities, 7 and for a passive body in a flow due to fluid mediated interactions of a nearby actively controlled object. 8 Here, we consider a simplified concept for locomotion formed by two steadily counter-rotating cylinders that are linked together by their centers of mass. This simple system is characterized by two parameters, the non-dimensional width and the rotational Reynolds number of the cylinders.…”

Reinforcement Learning and Wavelet Adapted Vortex Methods for Simulations of Self-propelled Swimmers

“…In this work, in order to develop simulations of fish schooling, the behavioral policies of multiple self-propelled two-dimensional swimmers are identified by coupling a reinforcement learning (RL) technique to a multiresolution remeshed vortex method with Brinkman penalization and projection approach [30,22]. To the best of our current knowledge, this coupling of RL algorithms and simulations of the NavierStokes equations has never been attempted before.…”

“…However, inviscid models do not capture processes such as flow separation and dynamic interactions between vortices in fish wakes. Viscous effects have been shown [22] to dramatically affect not only the quantitative but also the qualitative response of a system of bodies interacting in a flow field.…”