This project implements a high-performance, multiphysics Computational Fluid Dynamics (CFD) engine designed for modern web browsers. It utilizes the Lattice Boltzmann Method (LBM) to achieve real-time interactivity by leveraging C++ compiled to WebAssembly (WASM), SIMD vectorization, and multi-threaded execution.
For an in-depth exploration of the mathematical derivations and implementation details, refer to the included technical paper: Turbulence-sim.pdf.
The simulation framework enables the study of complex fluid behavior at interactive frame rates. Unlike traditional Navier-Stokes solvers, the LBM approaches fluid dynamics from a mesoscopic kinetic theory perspective, making it highly suitable for parallel execution on commodity hardware.
- Turbulence Modeling: Large Eddy Simulation (LES) using the Smagorinsky subgrid-scale model to resolve high Reynolds number flows.
- Thermodynamics: Advection-diffusion of temperature coupled with the momentum equations via the Boussinesq approximation for buoyancy-driven flows.
- Non-Newtonian Rheology: Implementation of the Ostwald-de Waele power-law model for shear-thinning and shear-thickening fluids.
- Multiphase Interactions: Surface tension and phase separation modeled via the Shan-Chen pseudopotential method.
- Porous Media: Darcy-Brinkman-Forchheimer drag terms for simulating flow through permeable structures.
- Stability Enhancements: Back and Forth Error Compensation and Correction (BFECC) for scalar advection and vorticity confinement to preserve small-scale eddies.
- Engine: C++17 implementation of the D2Q9 lattice model.
- Optimization: 128-bit WASM SIMD intrinsics for vectorized collision and streaming steps.
- Parallelism: Multi-threaded domain decomposition using
pthreads(compiled to Web Workers). - Memory Management: Direct manipulation of the WASM linear heap to minimize data transfer overhead between the physics engine and JavaScript.
- GPU Acceleration: Field visualization (vorticity, velocity, density, pressure) processed via fragment shaders.
- Particle System: Lagrangian particle tracers (up to 10,000,000) managed entirely on the GPU using Transform Feedback.
- Visualization: Support for multiple HDR colormaps (Turbo, Viridis, Inferno) and post-processing filters (Gaussian blur, edge detection).
| Directory/File | Description |
|---|---|
src/ |
C++ source code for the fluid engine and headers. |
web/ |
Target directory for compiled WASM, HTML, and JS assets. |
main.js |
Simulation orchestration and UI management. |
renderer.js |
WebGL2 context and particle system implementation. |
shaders.js |
GLSL shader sources for visualization and GPU computing. |
launch.bat |
Automated build and development server script for Windows. |
Makefile |
Build configuration for Unix-like systems. |
Turbulence-sim.pdf |
Comprehensive mathematical and technical documentation. |
- Emscripten SDK (emsdk): For C++ to WebAssembly compilation.
- Python 3: To serve the application with required security headers (
COOP/COEP).
The launch.bat script automates the compilation, asset synchronization, and server initialization.
launch.batUse the provided Makefile to compile the engine:
make build
# Serve the 'web' directory using the provided python script
python3 server.py 8005 webThis simulation requires SharedArrayBuffer for multithreading. Your web server must provide the following headers for the simulation to initialize:
Cross-Origin-Opener-Policy: same-originCross-Origin-Embedder-Policy: require-corp
The included server.py and launch.bat handle this automatically for local development.
- Interaction: Use the mouse or touch input to inject momentum, density, or temperature.
- Presets: Select from various physical configurations (e.g., Molten Gold, Superfluid Helium) via the GUI.
- Parameters: Real-time adjustment of viscosity, gravity, thermal expansion, and resolution.
- Shortcuts:
Space: Toggle pause.R: Reset simulation.P: Toggle particle visibility.
Licensed under the MIT License.