Optimal wavy surface to suppress vortex shedding using second-order sensitivity to shape changes

TL;DR

This paper introduces a computational method to identify optimal second-order shape modifications, specifically spanwise-wavy surfaces, that effectively suppress vortex shedding in cylinder wakes, validated through stability analysis at different Reynolds numbers.

Contribution

It extends sensitivity analysis to second-order for shape optimization in flow control, enabling the design of optimal surface shapes to stabilize flow instabilities.

Findings

Optimal wavy surface suppresses wake instability at Re=50 and Re=100.

Maximum height of 1% of cylinder diameter is sufficient for stabilization at Re=100.

Optimal shape passively creates streaks that alter wall velocity components, stabilizing flow.

Abstract

A method to find optimal 2nd-order perturbations is presented, and applied to find the optimal spanwise-wavy surface for suppression of cylinder wake instability. Second-order perturbations are required to capture the stabilizing effect of spanwise waviness, which is ignored by standard adjoint-based sensitivity analyses. Here, previous methods are extended so that (i) 2nd-order sensitivity is formulated for base flow changes satisfying linearised Navier-Stokes, and (ii) the resulting method is applicable to a 2D global instability problem. This makes it possible to formulate 2nd-order sensitivity to shape modifications. Using this formulation, we find the optimal shape to suppress the a cylinder wake instability. The optimal shape is then perturbed by random distributions in full 3D stability analysis to confirm that it is a local optimal at the given amplitude and wavelength. Furthermore, it is shown that none of the 10 random wavy shapes alone stabilize the wake flow at Re=50, while the optimal shape does. At Re=100, surface waviness of maximum height 1% of the cylinder diameter is sufficient to stabilize the flow. The optimal surface creates streaks by passively extracting energy from the base flow derivatives and effectively altering the tangential velocity component at the wall, as opposed to spanwise-wavy suction which inputs energy to the normal velocity component at the wall. This paper presents a fully two-dimensional and computationally affordable method to find optimal 2nd-order perturbations of generic flow instability problems and any boundary control (such as boundary forcing, shape modulation or suction).