JANC: A cost-effective, differentiable compressible reacting flow solver featured with JAX-based adaptive mesh refinement

JANC, as the abbreviation for “JAX-AMR & Combustion”, is a fully-differentiable compressible reacting flow solver based on JAX-AMR.

Authors:

• Haocheng Wen
• Faxuan Luo

Basic features of JANC

• Conjunction with JAX-AMR, allowing cost-effective large-scale simulations.
• Adoption of structured Cartesian grid, dimensionless equations, high-order finite difference method, point-implicit chemical source advancing in the solver.
• Inheriting the basic features of JAX, including fully differentiable, compatible with CPUs/GPUs/TPUs computation, and convenient parallel management.
• Programmed by Python, allowing rapid and efficient prototyping of projects.

Physical models and numerical methods

• Adaptive mesh refinenment (JAX-AMR)
• Dimensionless computation
• Explicit time advancing (RK3)
• High-order spatial reconstruction (WENO-5)
• Riemann solvers (Lax-Friedrichs)
• Point-implicit chemical source advancing
• CPU/GPU/TPU capability
• Parallel computation on GPU/TPU (only for the core solver in current version)

For the details, please refer to our paper.

Quick Installation

JANC CFD solver and JAX-AMR amr capabilities can be easily installed using pip install git:

pip install git+https://github.com/JA4S/JANC.git

Example

Rotating detonation combustor (RDC) simulation on 1,600,000 grids with 9sp-19r-H2-Air detailed reaction achieved within 45 minutes on single A100 GPU.

Open janc_basic_example1.ipynb in Google Colab to run the example.

Detonation tube simulation on 4,000,000 grids with 9sp-19r-H2-Air detailed reaction achieved within 1 hour on single A100 GPU.

Open janc_basic_example2.ipynb in Google Colab to run the example.

State of the Project

• 2D solver for Euler equations ✅
• conjuction with the CFD solver ✅
• Parallel compuation for the core solver ✅
• 3D solver for Navier-Stocks equations (soon)
• Implicit time advancing (soon)
• Turbulence model
• DPM model based on Euler-lagrange method
• Mixing-precision computation
• Parallel computation with JAX-AMR

Citation

JANC: A cost-effective, differentiable compressible reacting flow solver featured with JAX-based adaptive mesh refinement

@article{Wen2025,
   author = {Haocheng Wen and Faxuan Luo and Sheng Xu and Bing Wang},
   doi = {10.48550/arXiv.2504.13750},
   journal = {arXiv preprint},
   title = {JANC: A cost-effective, differentiable compressible reacting flow solver featured with JAX-based adaptive mesh refinement},
   year = {2025}
}

License

This project is licensed under the MIT License - see the LICENSE file or for details https://en.wikipedia.org/wiki/MIT_License.