This project was developed as part of the 18.337 class project.
This project leverages the Julia scientific machine learning ecosystem to establish a robust, high-performance simulation platform for the Cahn-Hilliard phase field model, a method used in studying phase-separating electrode materials. The work addresses common numerical challenges associated with simulating the stiff, nonlinear Cahn-Hilliard partial differential equation in custom geometries.
The Cahn-Hilliard model is a powerful tool for simulating dynamics in lithium ion battery materials. However, this model presents numerous numerical challenges due to its complexity. By leveraging Julia's high-performance scientific computing suite, this project provides a robust and efficient solution for these challenges.
- High-performance simulations of the Cahn-Hilliard phase field model
- Support for GPU parallelization
- Applications to PDE-constrained optimization
To see example usage of this model, please checkout the jupyter notebooks / julia files in the benchmark folder.
This project lays the foundation for future studies on learning constitutive relationships in batteries at the population scale. We have planned several improvements and novel applications to further enhance the functionality and use-cases of the platform.
- Samuel Degnan-Morgenstern stm16109@mit.edu
- Bazant Group - MIT Dept. of Chemical Engineering