This package implements energy-conserving solvers for the incompressible Navier-Stokes equations on a staggered Cartesian grid. It is based on the Matlab package INS2D/INS3D. The simulations can be run on the single/multithreaded CPUs or Nvidia GPUs.
| Documentation | Workflows | Code coverage | Quality assurance |
|---|---|---|---|
  |  |  |  |
This package implements energy-conserving solvers for the incompressible Navier-Stokes equations on a staggered Cartesian grid. It is based on the Matlab package INS2D/INS3D. The simulations can be run on the single/multithreaded CPUs or Nvidia GPUs.
This package also provides experimental support for neural closure models for large eddy simulation.
To install IncompressibleNavierStokes, open up a Julia-REPL, type ] to get
into Pkg-mode, and type:
(v1.10) pkg> add IncompressibleNavierStokes
which will install the package and all dependencies to your local environment.
Note that IncompressibleNavierStokes requires Julia version 1.9 or above.
See the
Documentation
for examples of some typical workflows. More examples can be found in the
examples directory.
See here for the source code used in the paper Discretize first, filter next: learning divergence-consistent closure models for large-eddy simulation.
The velocity and pressure fields may be visualized in a live session using
Makie. Alternatively,
ParaView may be used, after exporting individual
snapshot files using the save_vtk function, or the full time series using the
VTKWriter processor.
Make sure to have the GLMakie and IncompressibleNavierStokes installed:
using Pkg
Pkg.add(["GLMakie", "IncompressibleNavierStokes"])
Then run run the following code to make a short animation:
using GLMakie
using IncompressibleNavierStokes
# Setup
setup = Setup(
x = (tanh_grid(0.0, 2.0, 200, 1.2), tanh_grid(0.0, 1.0, 100, 1.2)),
boundary_conditions = ((DirichletBC(), DirichletBC()), (DirichletBC(), DirichletBC())),
temperature = temperature_equation(;
Pr = 0.71,
Ra = 1e7,
Ge = 1.0,
boundary_conditions = (
(SymmetricBC(), SymmetricBC()),
(DirichletBC(1.0), DirichletBC(0.0)),
),
),
)
# Solve equation
solve_unsteady(;
setup,
ustart = velocityfield(setup, (dim, x, y) -> zero(x)),
tempstart = temperaturefield(setup, (x, y) -> 1 / 2 + sinpi(30 * x) / 100),
tlims = (0.0, 30.0),
Δt = 0.02,
processors = (;
anim = animator(;
setup,
path = "temperature.mp4",
fieldname = :temperature,
colorrange = (0.0, 1.0),
size = (900, 500),
colormap = :seaborn_icefire_gradient,
nupdate = 5,
),
),
)
Discovering deep physics models with differentiable programming
CoupledNODE.jl is a SciML repository that extends NODEs (Neural Ordinary Differential Equations) to C-NODEs (Coupled Neural ODEs), providing a data-driven approach to modelling solutions for multiscale systems when exact solutions are not feasible.
