Adjoint-based iterative methods are employed in order to compute
optimal disturbances in the case of both linear and non-linear
perturbations in shear flows. The Lagrangian approach is used where an
objective function is chosen and constraints are assigned. We are
looking for stationary points of the Lagrange function with respect to
the different design variables where optimality is fulfilled. To this
purpose power iterations within a matrix-free framework are used, where the
state is marched forward in time with a normal DNS solver and backward
with the adjoint solver until a chosen convergence criterion is fulfilled.
In the linear framework the global linear stability of the flat-plate
boundary-layer
flow to three-dimensional disturbances is studied. We consider both the
optimal
initial condition leading to the largest departure from the laminar flow
at finite
times and the optimal time-periodic forcing leading to the largest
asymptotic
response. The objective function of the optimization is the kinetic
energy of the
flow perturbations and the constraints involve the linearised
Navier--Stokes equations.
In the non-linear framework, we determine the initial condition on the
laminar/turbulent boundary closest to the laminar state for plane
Couette flow. Resorting to the general evolution criterion of
non-equilibrium systems we optimize the route to the statistically
steady turbulent state, i.e. the state characterized by the largest
entropy production. This is the first time information from the
fully turbulent state is included in the optimisation procedure.
We demonstrate that the optimal initial condition is localized in
space for realistic flow domains, larger than those previously used.
We investigate the transition path and show how localized perturbations
evolve into bent streaks that later break down to turbulence.
|