| 1 |
In this demo we will use goal oriented adaptivity and error control |
| 2 |
which applies a duality technique to derive error estimates taken |
| 3 |
directly from the computed solution which then are used to weight |
| 4 |
local residuals. To this end, we derive an :math:`\textit{a |
| 5 |
posteriori}` error estimate and error indicators. We define a goal |
| 6 |
functional :math:`\mathcal{M} : V \rightarrow \mathbb{R}`, which |
| 7 |
expresses the localized physical properties of the solution of a |
| 8 |
simulation. The objective of goal oriented adaptive error control is |
| 9 |
to minimize computational work to obtain a given level of accuracy in |
| 10 |
:math:`\mathcal{M}`. |
| 11 |
|
| 12 |
We will thus illustrate how to: |
| 13 |
|
| 14 |
* Solve a linear partial differential equation with automatic adaptive mesh refinement |
| 15 |
* Define a goal functional |
| 16 |
* Use :py:class:`AdaptiveLinearVariationalSolver <dolfin.cpp.fem.AdaptiveLinearVariationalSolver>` |
| 17 |
|
| 18 |
The two solutions for u in this demo will look as follows, where the |
| 19 |
first is the unrefined while the second is the refined solution: |
| 20 |
|
| 21 |
.. image:: ../u_unrefined.png |
| 22 |
:scale: 75 % |
| 23 |
|
| 24 |
.. image:: ../u_refined.png |
| 25 |
:scale: 75 % |
| 26 |
|
| 27 |
|
| 28 |
Equation and problem definition |
| 29 |
------------------------------- |
| 30 |
|
| 31 |
The Poisson equation is the canonical elliptic partial differential |
| 32 |
equation. For a domain :math:`\Omega \subset \mathbb{R}^n` with |
| 33 |
boundary :math:`\partial \Omega = \Gamma_{D} \cup \Gamma_{N}`, the |
| 34 |
Poisson equation with particular boundary conditions reads: |
| 35 |
|
| 36 |
|
| 37 |
.. math:: |
| 38 |
|
| 39 |
- \nabla^{2} u &= f \quad {\rm in} \ \Omega, \\ |
| 40 |
u &= 0 \quad {\rm on} \ \Gamma_{D}, \\ |
| 41 |
\nabla u \cdot n &= g \quad {\rm on} \ \Gamma_{N}. \\ |
| 42 |
|
| 43 |
Here, :math:`f` and :math:`g` are input data and n denotes the outward |
| 44 |
directed boundary normal. The variational form of Poisson equation |
| 45 |
reads: find :math:`u \in V` such that |
| 46 |
|
| 47 |
.. math:: |
| 48 |
|
| 49 |
a(u, v) = L(v) \quad \forall \ v \in \hat{V}, |
| 50 |
|
| 51 |
which we will call the continous primal problem, where :math:`V`, |
| 52 |
:math:`\hat{V}` are the trial- and test spaces and |
| 53 |
|
| 54 |
|
| 55 |
.. math:: |
| 56 |
|
| 57 |
a(u, v) &= \int_{\Omega} \nabla u \cdot \nabla v \, {\rm d} x, \\ |
| 58 |
L(v) &= \int_{\Omega} f v \, {\rm d} x + \int_{\Gamma_{N}} g v \, {\rm d} s. |
| 59 |
|
| 60 |
The expression :math:`a(u, v)` is the bilinear form and :math:`L(v)` |
| 61 |
is the linear form. It is assumed that all functions in :math:`V` |
| 62 |
satisfy the Dirichlet boundary conditions (:math:`u = 0 \ {\rm on} \ |
| 63 |
\Gamma_{D}`). |
| 64 |
|
| 65 |
The above definitions is that of the continuous problem. In the actual |
| 66 |
computer implementation we use a descrete formulation which reads: |
| 67 |
find :math:`u \in V_h` such that |
| 68 |
|
| 69 |
|
| 70 |
.. math:: |
| 71 |
|
| 72 |
a(u_h, v) = L(v) \quad \forall \ v \in \hat{V}_h. |
| 73 |
|
| 74 |
We will refer to the above equation as the discrete primal |
| 75 |
problem. Here, :math:`V_h` and :math:`\hat{V_h}` are finite |
| 76 |
dimensional subspaces. |
| 77 |
|
| 78 |
The weak residual is defined as |
| 79 |
|
| 80 |
.. math:: |
| 81 |
|
| 82 |
r(v) = L(v) - a(u_h, v). |
| 83 |
|
| 84 |
By the Galerkin orthogonality, we have |
| 85 |
|
| 86 |
|
| 87 |
.. math:: |
| 88 |
|
| 89 |
r(v) = L(v) - a(u_h, v) = a(u_h, v) - a(u_h, v) = 0\,\, \forall v \in \hat{V}_{h}, |
| 90 |
|
| 91 |
which means that the residual vanishes for all functions in |
| 92 |
:math:`\hat{V}_{h}`. This property is used further in the derivation |
| 93 |
of the error estimates. wish to compute a solution :math:`u_h` to the |
| 94 |
discrete primal problem defined above such that for a given tolerance |
| 95 |
:math:`\mathrm{TOL}` we have |
| 96 |
|
| 97 |
.. math:: |
| 98 |
|
| 99 |
\eta = \left| \mathcal{M}(u_h) - \mathcal{u_h} \right| \leq \mathrm{TOL}. |
| 100 |
|
| 101 |
Next we derive an :math:`\textit{a posteriori}` error estimate by |
| 102 |
defining the discrete dual variational problem: find :math:`z_h \in |
| 103 |
V_h^*` such that |
| 104 |
|
| 105 |
.. :math:: |
| 106 |
|
| 107 |
a^*(z_h,v) = \mathcal{v}, \quad \forall v \in \hat{V}_h^*. |
| 108 |
|
| 109 |
Here :math:`V^*, \hat{V}_h^*` are the dual trial- and test spaces and |
| 110 |
:math:`a^* : V^* \times \hat{V}^* \rightarrow \mathbb{R}` is the |
| 111 |
adjoint of :math:`a` such that :math:`a^*(v,w) = a(w,v)`. We find that |
| 112 |
|
| 113 |
.. math:: |
| 114 |
|
| 115 |
\mathcal{M}(u - \mathcal{u_h}) = a^*(z, u - u_h) = a(u - u_h, z) = a(u,z) - a(u_h,z) = L(z) - a(u_h,z) = r(z) |
| 116 |
|
| 117 |
and by Galerkin orthogonality we have :math:`r(z) = r(z - v_h)\,\, |
| 118 |
\forall v_h \in \hat{V}_h`. Note that the residual vanishes if |
| 119 |
:math:`z \in \hat{V}_h^*` and has to either be approximated in a |
| 120 |
higher order element space or one may use an extrapolation. The choice |
| 121 |
of goal functional depends on what quantity you are interested in. |
| 122 |
Here, we take the goal functional to be defined as |
| 123 |
|
| 124 |
.. math:: |
| 125 |
|
| 126 |
\mathcal{M}(u) = \int_{\Omega} u dx. |
| 127 |
|
| 128 |
|
| 129 |
We use :math:`D\ddot{o}rfler` marking as the mesh marking procedure. |
| 130 |
|
| 131 |
In this demo, we shall consider the following definitions of the input functions, the domain, and the boundaries: |
| 132 |
|
| 133 |
* :math:`\Omega = [0,1] \times [0,1]\,` (a unit square) |
| 134 |
* :math:`\Gamma_{D} = \{(0, y) \cup (1, y) \subset \partial \Omega\}\,` (Dirichlet boundary) |
| 135 |
* :math:`\Gamma_{N} = \{(x, 0) \cup (x, 1) \subset \partial \Omega\}\,` (Neumann boundary) |
| 136 |
* :math:`g = \sin(5x)\,` (normal derivative) |
| 137 |
* :math:`f = 10\exp(-((x - 0.5)^2 + (y - 0.5)^2) / 0.02)\,` (source term) |
| 138 |
|