CFDeez is a 2D finite-volume computational fluid dynamics (CFD) application and library, developed entirely from scratch. While primarily created as a learning tool, it can also serve as a helpful resource for those interested in CFD.

Velocity profile around an object.

CFD has a reputation of being difficult. So the goal was to see if 2 univeristy students with big egos but no real understanding of anything could pull it off. Development is still underway so our egos are in a superposition of being massive and also heavily bruised.

To get started with CFDeez, you’ll need the Odin compiler installed, as no pre-built binaries will be provided for early versions.

# Clone the repository
git clone https://github.com/Rwn-A/cfdeez
cd cfdeez

# Build the CLI tool (optimizations really help)
odin build src/app/cli -out:cfdeez.exe -o:speed 
# add -define:LOG_LINSOLVE_PROGRESS=true to see how the linear solvers are progressing if you run into non-convergence.

# Run a basic example
./cfdeez.exe ./example/00_basic.fml

After this results will be saved to a .out directory located beside your executable.

CFDeez is not designed as production CFD software. It is CPU only. While we might add GPU processing, it is not a guarantee and outside the scope of this project for now. The primary goal of this project is to understand how CFD works, and write readable code that represents the physics. Specific discretization schemes are currently not configurable.

Features:

• ✅ 2D incompressible flow
• ✅ 2D transport
• ✅ Unstructured meshes
• ✅ Transient and steady-state
• ✅ Output to VTU and CSV
• ✅ Configurable

Currently, we have no 1st party post-processing tools, best bet is to use VTK output option and use something like ParaView, or pyvista with python. You can also output to CSV and write a custom visualization.

In no particular order, the below features are things we are interested in but have no current plans to attempt to implement.

• Turbulence Modelling
• 3D
• 1st Party Mesh Generation
• Compressible Flow
• Configurable schemes
• Native GUI application

The best way to learn the configuration options is to review the examples directory. Each example is numbered in order of complexity. There are also some prebuilt .msh files available.

A configuration consists of two parts.

1. .fml file - Most of the configuration options are here, this is the file you pass to the executable.
2. .msh file - this defines the mesh for your simulation, the path to this file is referenced in the main config file.

To build a .msh file gmsh is required. Or you can convert from another format using something like meshio.

• Units are SI, this project is proudly Canadian 🍁, no freedom units.
• ✓* indicates that the section is required if the parent section is included
• ✓** indicates that the section is required based on the presence of other sections. The application will tell you if your missing an option.
• To provide initial conditions from a file, provide the path as a string (this feature is currently unimplemented.)
• To provide inflow profile & initial conditions based on mesh coordinates, you can use an expression.
  Such as inflow_profile: <x + cos(y)>, the basic math functions are supported, along with x and y to capture coordinates.
• To provide a constant value for initial conditions & inflow, you can provide a number directly.

CFDeez is comprised of 3 components.

1. CFD Library /src/cfd
   • The core cfd code is its own library, it includes code for the fields, discretizations, linear systems and more
   • Includes optional sub-packages for outputting fields, reading .msh files, and pre-built solver algorithms.
2. Configuration Language /src/fml
   • CFDeez uses a custom JSON esque configuration language called FML, FML can be used standalone although I can't think of many use cases outside of this application.
3. Application Interface /src/app
   • The configuration language and CFD library are wrapped together into a single executable with a predefined configuration schema.

This is an engineering simulation in spirit so we have tried to use accurate methods. That being said, most discretizations are only first-order accurate, but implicit for stability. Higher order methods may be implemented in the future.

So far, we have only verified laminar channel flow, backward-facing step also passed the eye-test.

We expect fully developed flow to be about 1.5 times the inlet velocity at most, and parabolic about the y-axis, below is the graph generated by running the channel_flow.py script in the validation directory.

The CFD library and application code are licensed under the GPLv3 license. The FML configuration language is licensed separately under the MIT license.

Any and all contributions welcome, that does not mean any and all contributions will be accepted.

• [1] H. K. Versteeg and W. Malalasekera, An Introduction to Computational Fluid Dynamics: The Finite Volume Method, 2nd ed. Harlow, England: Pearson Education, 2007.

• [2] J. H. Ferziger, M. Perić, and R. L. Street, Computational Methods for Fluid Dynamics, 4th ed. Cham, Switzerland: Springer, 2020.

• [3] "CFD Direct," CFD Direct Documentation, Accessed: Apr 18, 2025. [Online]. Available: https://doc.cfd.direct/

• [4] "OpenFOAM – Official home of The Open Source CFD Toolbox," OpenFOAM Foundation, Accessed: Jun. 18, 2025. [Online]. Available: https://openfoam.org/

• [5] Fluid Mechanics 101, YouTube, [Online]. Available: https://www.youtube.com/@fluidmechanics101