Sensitivity & Flow Control

We study flow structures using a variety of techniques (e.g. Global stability analysis, Proper Orthogonal Decomposition, Dynamic Mode Decomposition, Machine Learning & Neural Networks) to understand physics of unsteady complex flows. These techniques allow to compute gradients and sensitivities of the flow to local forcing or geometrical changes, that we use for flow control. Coupling these sensitivities to optimisers, we can obtain new geometries that inhibit undesired unsteady flow features to control unsteady loads or minimise acoustic emissions.

Fig 1: Direct and Adjoint modes for a circular cylinder at Re=45

Fig 2: Sensitivity for a circular cylinder at Re=45: (a) Structural sensitivity and (b) Sensitivity to base flow modification

Fig 3: 3D global instability analysis of an L-shaped cavity at Reynolds 900 and non-dimensional spanwise length of 0.628, simulation with polynomial order k=17 in the x-y plane and 64 Fourier planes in the z-direction: (a) Nonlinear DNS (N-DNS): 10 iso-surfaces of velocity, (b) Linearised DNS (LDNS): 10 iso-surfaces of w-velocity, (c) N-DNS+DMD: real part of the most unstable mode and (d) LDNS+DMD: real part of the most unstable mode

Fig 4: Direct Numerical Simulation of a lid-driven L-shaped cavity and onset of three dimensional instabilities. a) 2D base flow and b) temporal evolution of W-velocity. The curve shows exponential growth (linearised regime) and saturation (non-linear regime).

Fig 5: Structural sensitivity for a turbulent NACA0012 airfoil at transonic conditions: Study of the buffet phenomenon

Fig 6: Dynamic Mode Decomposition of the 3D ILES computation for a turbulent flow over a cavity Re = 8.6x10⁵ and Mach 0.8. Left: DMD mode at 2000 Hz and Right: DMD mode at 4000 Hz

Fig 7: Linear stability analysis of a submerged cylinder