DeepPhysicsSim aims to provide a physics-based simulation of fluid flow and temperature distribution in a 3D domain using deep learning. Built on PyTorch, this project incorporates various physical laws and constraints into the machine learning model, effectively turning the neural network into a "differentiable physics simulator".
- Physics-based loss functions incorporating adiabatic residuals, temperature residuals, fluid residuals, and more.
- 3D scatter plot visualizations for in-depth analysis.
- Utilizes efficient optimization algorithms for faster convergence.
- Offers flexibility to adapt to different physical scenarios.
- Python 3.x
- PyTorch
- Pandas
- NumPy
- Matplotlib
git clone https://github.com/GiulioMa/DeepPhysicsSim.git
cd DeepPhysicsSim
pip install -r requirements.txt