A Penalized Neumann Control Approach for Solving an Optimal Dirichlet Control Problem for the Navier--Stokes Equations (Q4210187)

The problem is that of minimizing the vorticity of viscous, incompressible flow in a bounded two- or three-dimensional domain \(\Omega\) with boundary \(\Gamma\) by controlling the boundary velocity. This means minimizing \[ {\mathcal J}({\mathbf u}, {\mathbf g}) = {\alpha \over 2} \int_\Omega | \text{curl} {\mathbf u}| ^2 d{\mathbf x} + {\beta \over 2} \int_\Gamma | {\mathbf g}| ^2 d\sigma \] \((\alpha, \beta > 0)\) among solutions of the steady state Navier-Stokes equations \[ - \nu \Delta {\mathbf u} + ({\mathbf u} \cdot \nabla){\mathbf u} + \nabla p = {\mathbf f}, \quad \text{div} {\mathbf u} = 0 \qquad (x \in \Omega), \] satisfying the Dirichlet boundary condition \[ {\mathbf u} = {\mathbf g} \quad (x \in \Gamma), \] where \({\mathbf f}\) is a prescribed forcing term and the boundary velocity \({\mathbf g}\) is the control. The motivation is: as irrotational flows have no local flow recirculation, it is expected that minimizing the vorticity will reduce flow recirculation. The authors approach this problem by penalization: the boundary condition is approximated by the nonlinear Robin boundary condition \[ - p {\mathbf n} + \nu {\partial {\mathbf u} \over \partial {\mathbf n}} - {1 \over 2} ({\mathbf u}\cdot {\mathbf n}) {\mathbf u} + {1 \over \varepsilon}{\mathbf u} = {1 \over \varepsilon}{\mathbf g} \quad (x \in \Gamma), \] as \(\varepsilon \to 0.\) The solution of the \(\varepsilon\)-problem is shown to exist and to converge (for a sequence \(\{\varepsilon_n\})\) to the solution of the original problem. The method is then applied to two examples where \(\Omega\) is the unit square. The computation is performed by means of a finite element approximation, and the results confirm the expected reduction in recirculation.