| 1 |
.. Documentation for the incompressible Navier-Stokes demo from DOLFIN. |
| 2 |
|
| 3 |
.. _demo_pde_navier_stokes_python_documentation: |
| 4 |
|
| 5 |
Incompressible Navier-Stokes equations |
| 6 |
====================================== |
| 7 |
|
| 8 |
This demo is implemented in a single Python file, |
| 9 |
:download:`demo_navier-stokes.py`, which contains both the variational |
| 10 |
forms and the solver. |
| 11 |
|
| 12 |
.. include:: ../common.txt |
| 13 |
|
| 14 |
Implementation |
| 15 |
-------------- |
| 16 |
|
| 17 |
This demo is implemented in the :download:`demo_navier-stokes.py` file. |
| 18 |
|
| 19 |
First, the :py:mod:`dolfin` module is imported: |
| 20 |
|
| 21 |
.. code-block:: python |
| 22 |
|
| 23 |
from dolfin import * |
| 24 |
|
| 25 |
For the parallel case, we turn off log messages from processes other than |
| 26 |
the the root process to avoid excessive output: |
| 27 |
|
| 28 |
.. code-block:: python |
| 29 |
|
| 30 |
# Print log messages only from the root process in parallel |
| 31 |
parameters["std_out_all_processes"] = False; |
| 32 |
|
| 33 |
We then load the mesh for the L-shaped domain from file: |
| 34 |
|
| 35 |
.. code-block:: python |
| 36 |
|
| 37 |
# Load mesh from file |
| 38 |
mesh = Mesh("../lshape.xml.gz") |
| 39 |
|
| 40 |
We next define a pair of function spaces :math:`V` and :math:`Q` for |
| 41 |
the velocity and pressure, and trial and test functions on these |
| 42 |
spaces: |
| 43 |
|
| 44 |
.. code-block:: python |
| 45 |
|
| 46 |
# Define function spaces (P2-P1) |
| 47 |
V = VectorFunctionSpace(mesh, "Lagrange", 2) |
| 48 |
Q = FunctionSpace(mesh, "Lagrange", 1) |
| 49 |
|
| 50 |
# Define trial and test functions |
| 51 |
u = TrialFunction(V) |
| 52 |
p = TrialFunction(Q) |
| 53 |
v = TestFunction(V) |
| 54 |
q = TestFunction(Q) |
| 55 |
|
| 56 |
The time step, the length of the time interval, and the kinematic |
| 57 |
viscosity are defined by: |
| 58 |
|
| 59 |
.. code-block:: python |
| 60 |
|
| 61 |
# Set parameter values |
| 62 |
dt = 0.01 |
| 63 |
T = 3 |
| 64 |
nu = 0.01 |
| 65 |
|
| 66 |
The time-dependent pressure boundary condition can be defined using |
| 67 |
the :py:class:`Expression <dolfin.functions.expression.Expression>` |
| 68 |
class: |
| 69 |
|
| 70 |
.. code-block:: python |
| 71 |
|
| 72 |
# Define time-dependent pressure boundary condition |
| 73 |
p_in = Expression("sin(3.0*t)", t=0.0, degree=2) |
| 74 |
|
| 75 |
Note that the variable ``t`` is not automatically updated during |
| 76 |
time-stepping, so we must remember to manually update the value of the |
| 77 |
current time in each time step. |
| 78 |
|
| 79 |
We may now define the boundary conditions for the velocity and |
| 80 |
pressure. We define one no-slip boundary condition for the velocity |
| 81 |
and a pair of boundary conditions for the pressure at the inflow and |
| 82 |
outflow boundaries: |
| 83 |
|
| 84 |
.. code-block:: python |
| 85 |
|
| 86 |
# Define boundary conditions |
| 87 |
noslip = DirichletBC(V, (0, 0), |
| 88 |
"on_boundary && \ |
| 89 |
(x[0] < DOLFIN_EPS | x[1] < DOLFIN_EPS | \ |
| 90 |
(x[0] > 0.5 - DOLFIN_EPS && x[1] > 0.5 - DOLFIN_EPS))") |
| 91 |
inflow = DirichletBC(Q, p_in, "x[1] > 1.0 - DOLFIN_EPS") |
| 92 |
outflow = DirichletBC(Q, 0, "x[0] > 1.0 - DOLFIN_EPS") |
| 93 |
bcu = [noslip] |
| 94 |
bcp = [inflow, outflow] |
| 95 |
|
| 96 |
|
| 97 |
We collect the boundary conditions in the two lists ``bcu`` and |
| 98 |
``bcp`` so that we may easily iterate over them below when we apply |
| 99 |
the boundary conditions. This makes it easy to add new boundary |
| 100 |
conditions or use this demo program to solve the Navier-Stokes |
| 101 |
equations on other geometries. |
| 102 |
|
| 103 |
We next define the functions and the coefficients that will be used |
| 104 |
below: |
| 105 |
|
| 106 |
.. code-block:: python |
| 107 |
|
| 108 |
# Create functions |
| 109 |
u0 = Function(V) |
| 110 |
u1 = Function(V) |
| 111 |
p1 = Function(Q) |
| 112 |
|
| 113 |
# Define coefficients |
| 114 |
k = Constant(dt) |
| 115 |
f = Constant((0, 0)) |
| 116 |
|
| 117 |
Note that one may use the time step ``dt`` directly in the |
| 118 |
form. However, by using the :py:class:`Constant |
| 119 |
<dolfin.functions.constant.Constant>` class, we may freely change the |
| 120 |
size of the time step without triggering regeneration of code. |
| 121 |
|
| 122 |
The next step is now to define the variational problems for the three |
| 123 |
steps of Chorin's method. We do this by defining a pair of bilinear |
| 124 |
and linear forms for each step: |
| 125 |
|
| 126 |
.. code-block:: python |
| 127 |
|
| 128 |
# Tentative velocity step |
| 129 |
F1 = (1/k)*inner(u - u0, v)*dx + inner(grad(u0)*u0, v)*dx + \ |
| 130 |
nu*inner(grad(u), grad(v))*dx - inner(f, v)*dx |
| 131 |
a1 = lhs(F1) |
| 132 |
L1 = rhs(F1) |
| 133 |
|
| 134 |
# Pressure update |
| 135 |
a2 = inner(grad(p), grad(q))*dx |
| 136 |
L2 = -(1/k)*div(u1)*q*dx |
| 137 |
|
| 138 |
# Velocity update |
| 139 |
a3 = inner(u, v)*dx |
| 140 |
L3 = inner(u1, v)*dx - k*inner(grad(p1), v)*dx |
| 141 |
|
| 142 |
Since the bilinear forms do not depend on any coefficients that change |
| 143 |
during time-stepping, the corresponding matrices remain constant. We |
| 144 |
may therefore assemble these before the time-stepping begins: |
| 145 |
|
| 146 |
.. code-block:: python |
| 147 |
|
| 148 |
# Assemble matrices |
| 149 |
A1 = assemble(a1) |
| 150 |
A2 = assemble(a2) |
| 151 |
A3 = assemble(a3) |
| 152 |
|
| 153 |
# Use amg preconditioner if available |
| 154 |
prec = "amg" if has_krylov_solver_preconditioner("amg") else "default" |
| 155 |
|
| 156 |
During time-stepping, we will store the solution in VTK format |
| 157 |
(readable by MayaVi and Paraview). We therefore create a pair of files |
| 158 |
that can be used to store the solution. Specifying the ``.pvd`` suffix |
| 159 |
signals that the solution should be stored in VTK format: |
| 160 |
|
| 161 |
.. code-block:: python |
| 162 |
|
| 163 |
# Create files for storing solution |
| 164 |
ufile = File("results/velocity.pvd") |
| 165 |
pfile = File("results/pressure.pvd") |
| 166 |
|
| 167 |
The time-stepping loop is now implemented as follows: |
| 168 |
|
| 169 |
.. code-block:: python |
| 170 |
|
| 171 |
# Time-stepping |
| 172 |
t = dt |
| 173 |
while t < T + DOLFIN_EPS: |
| 174 |
|
| 175 |
# Update pressure boundary condition |
| 176 |
p_in.t = t |
| 177 |
|
| 178 |
We remember to update the current time for the time-dependent pressure |
| 179 |
boundary value. |
| 180 |
|
| 181 |
For each of the three steps of Chorin's method, we assemble the |
| 182 |
right-hand side, apply boundary conditions and solve a linear |
| 183 |
system. Note the different use of preconditioners. Incomplete LU |
| 184 |
factorization is used for the computation of the tentative velocity |
| 185 |
and the velocity update, while algebraic multigrid is used for the |
| 186 |
pressure equation if available: |
| 187 |
|
| 188 |
.. code-block:: python |
| 189 |
|
| 190 |
# Compute tentative velocity step |
| 191 |
begin("Computing tentative velocity") |
| 192 |
b1 = assemble(L1) |
| 193 |
[bc.apply(A1, b1) for bc in bcu] |
| 194 |
solve(A1, u1.vector(), b1, "bicgstab", "default") |
| 195 |
end() |
| 196 |
|
| 197 |
# Pressure correction |
| 198 |
begin("Computing pressure correction") |
| 199 |
b2 = assemble(L2) |
| 200 |
[bc.apply(A2, b2) for bc in bcp] |
| 201 |
[bc.apply(p1.vector()) for bc in bcp] |
| 202 |
solve(A2, p1.vector(), b2, "bicgstab", prec) |
| 203 |
end() |
| 204 |
|
| 205 |
# Velocity correction |
| 206 |
begin("Computing velocity correction") |
| 207 |
b3 = assemble(L3) |
| 208 |
[bc.apply(A3, b3) for bc in bcu] |
| 209 |
solve(A3, u1.vector(), b3, "bicgstab", "default") |
| 210 |
end() |
| 211 |
|
| 212 |
Note the use of ``begin`` and ``end``; these improve the readability |
| 213 |
of the output from the program by adding indentation to diagnostic |
| 214 |
messages. |
| 215 |
|
| 216 |
At the end of the time-stepping loop, we plot the solution, store the |
| 217 |
solution to file, and update values for the next time step: |
| 218 |
|
| 219 |
.. code-block:: python |
| 220 |
|
| 221 |
# Plot solution |
| 222 |
plot(p1, title="Pressure", rescale=True) |
| 223 |
plot(u1, title="Velocity", rescale=True) |
| 224 |
|
| 225 |
# Save to file |
| 226 |
ufile << u1 |
| 227 |
pfile << p1 |
| 228 |
|
| 229 |
# Move to next time step |
| 230 |
u0.assign(u1) |
| 231 |
t += dt |
| 232 |
|
| 233 |
Finally, we call the ``interactive`` function to signal that the plot |
| 234 |
window should be kept open and allow us to interact with (zoom, |
| 235 |
rotate) the solution. |
| 236 |
|
| 237 |
Complete code |
| 238 |
------------- |
| 239 |
|
| 240 |
.. literalinclude:: demo_navier-stokes.py |
| 241 |
:start-after: # Begin demo |