Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Project 09: Nonmatching Grid Control with Overlapping Meshes

University of Pisa

Goal

Use nonmatching meshes to control a PDE by prescribing values on an overlapping control grid that is independent of the state mesh.

This project is inspired by deal.II step-60.

Mathematical Formulation

Let Ω\Omega be the physical domain and let Ωc\Omega_c be an overlapping control domain with its own mesh. The state is defined on Ω\Omega, while the control uu is defined on Ωc\Omega_c.

Let

Π:Uh(Ωc)L2(Ω)\Pi:U_h(\Omega_c)\to L^2(\Omega)

be the transfer operator from the nonmatching control mesh to the state mesh. For example, Πu\Pi u may be the restriction or interpolation of the control onto the overlap.

Solve

min(y,u)12yydL2(Ω)2+α2uurefL2(Ωc)2\min_{(y,u)} \frac12\|y-y_d\|_{L^2(\Omega)}^2 + \frac\alpha2\|u-u_{\mathrm{ref}}\|_{L^2(\Omega_c)}^2

subject to

{Δy=Πu+fin Ω,y=0on Ω.\begin{cases} -\Delta y=\Pi u+f & \text{in }\Omega,\\ y=0 & \text{on }\partial\Omega. \end{cases}

The weak state equation is

Ωyvdx=Ω(Πu)vdx+Ωfvdx.\int_\Omega \nabla y\cdot\nabla v\,dx = \int_\Omega (\Pi u)v\,dx + \int_\Omega f v\,dx.

The adjoint equation is

Ωpvdx=Ω(yyd)vdx.\int_\Omega \nabla p\cdot\nabla v\,dx = \int_\Omega (y-y_d)v\,dx.

The control equation is

α(uuref)Πp=0in Uh(Ωc),\alpha(u-u_{\mathrm{ref}}) - \Pi^*p=0 \qquad\text{in }U_h(\Omega_c),

where Π\Pi^* is the adjoint transfer from the state mesh to the control mesh.

Implementation Hints

(Πu,p)L2(Ω)=(u,Πp)L2(Ωc).(\Pi u,p)_{L^2(\Omega)} = (u,\Pi^*p)_{L^2(\Omega_c)}.

Relevant Examples

Deliverables