This repo contains a differentiable physics-based simulation tool. It focuses on additive manufacturing and uses gradient-based methods to optimize the process. Current version is developed on top of Taichi, which showed better performance compared to TensorFlow, PyTroch, and Jax in our tests.
- Explore different automatic differentiation libraries
- Automatic mesh and toolpath loader
- Visualization tools using pyvista
- Upgrade for unstructured shape functions
- Add temperature-dependent material properties
- Validate results with benchmarks
Thermal simulation results during the build:
Laser power evolution during optimization process:
git clone https://github.com/mojtabamozaffar/differentiable-simulation-am
cd differentiable-simulation-am
pip install -r requirements.txt
Execute jupyter notebook cells sequentially, unless instructed in the comments.
This project is released under the MIT License.