Physics-Regulated Interpretable Machine Learning Microstructure Evolution (PRIMME)

Description:

Physics-Regularized Interpretable Machine Learning Microstructure Evolution (PRIMME): This code can be used to train and validate PRIMME neural network models for simulating isotropic microstructural grain growth.

To Access the sample Training Dataset from SPPARKS, you can download it from here, it should be placed inside of the /.PRIMME/data directory.

Paper on the Model

Usage

There are two ways to run the program:

Google Colab

See the following Colab link to run PRIMME remotely

Local GUI

Dependencies

• Python 3.9 - 3.12
• For rest, see requirements.txt

Installation

Clone this repository and create virtual environment:

pip install virtualenv # if not done so already

git clone https://github.com/EAGG-UF/PRIMME.git
cd PRIMME
python3.9 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Run the GUI Application for Training and Running PRIMME

cd PRIMME
python gui_application.py

Visuals

Isotropic Case

         

Training on mode filter(left), Training on MCP(mid) and Training on phase field (right).

Anisotropic Case

    

Training on mode filter(left) and Training on phase field (right).

Notes:

• This model is often trained of SPPARKS data, see its GitHub and Documentation for more information.

Contributors:

Weishi Yan, Joel Harley, Joseph Melville, Kristien Everett, Tian Zhihui, Lin Yang, Vishal Yadav, Michael Tonks, Amanda Krause, Gabriel Castejon, Manas Adepu.

Affiliation:

1. University of Florida, SmartDATA Lab, Department of Electrical and Computer Engineering
2. University of Florida, Tonks Research Group, Department of Materials Science and Engineering

Funding Sponsors:

U.S. Department of Energy, Office of Science, Basic Energy Sciences under Award #DE-SC0020384 U.S. Department of Defence through a Science, Mathematics, and Research for Transformation (SMART) scholarship