- Install basic libraries -
pip install matplotlib cantera tensorflow - Install the SL-GPS library -
pip install slgps - Test your installation by running any of the files in the tests folder
- The trained neural network is stored as .h5 file which can be accessed and utilized to produce reduced mechanisms for any given composition, temperature and pressure.
NOTE: The default neural network architecture is 16x8 (2 hidden layers). To change the neural network architecture, go to the file and edit the function spec_train according to your needs.
For questions and discussions please join : https://discord.com/channels/1333609076726431798/1333610748424880128
Please feel free to ask any questions related to SL-GPS there.
This repository contains the means to create a neural network architecture for dynamic chemistry reduction based on reduction results from Global Pathway Selection. The basic procedure is to first run adaptive GPS for 0D auto-ignition simulation so as to create a dataset. This dataset is later used for training the Artificial Neural Network (ANN). You can reach out to us at [email protected] (Rohit Mishra) or [email protected] (Aaron Nelson) for code issues, suggestions and/or pull requests. The step-by-step procedure to use this repository is as follows:
This code was developed entirely in Python 3. Dependent packages include Cantera 2, Tensorflow 2, pandas, sklearn, numpy, pickle, and networkx. Code for GPS has been copied from https://github.com/golsun/GPS and modified to work in Python 3.
- Mishra, R., Nelson, A., Jarrahbashi, D., "Adaptive global pathway selection using artificial neural networks: A-priori study", Combustion and Flame, 244 (2022) 112279 [link]
- X. Gao, S. Yang, W. Sun, "A global pathway selection algorithm for the reduction of detailed chemical kinetic mechanisms", Combustion and Flame, 167 (2016) 238-247 [link]