Skip to content

Latest commit

 

History

History
49 lines (34 loc) · 2.49 KB

README.md

File metadata and controls

49 lines (34 loc) · 2.49 KB
Raw

Kinetic-constrained neural ODE (KCNODE)

This repository supplements our Chemical Engineering Journal paper (doi:10.1016/j.cej.2023.146869) and the earlier ChemRXIV preprint (doi:10.26434/chemrxiv-2023-x39xt).

Contents of the repository

  • data_generator.py - Functions for generating the data for the numerical experiment
  • KCNODE.py - Classes and methods for building neural ODE models
  • Baseline model.ipynb - Example using the baseline model
  • KCNODE FT.ipynb.ipynb - Example using the KCNODE model for the process of CO2 hydrogenation to hydrocarbons via FT (real data)
  • KCNODE methanation.ipynb - Example using the KCNODE model for the process of CO2 hydrogenation to methane (numerical experiment)
  • Training.ipynb - Example demonstrating training of the neural ODE models
  • /trained_models - Directory with a trained model of neural ODE
    • baseline.pt - The baseline neural ODE model.
    • KCNODE_methanation.pt - The KCNODE model for CO2 hydrogenation to CH4
    • KCNODE_FT.pt - The KCNODE model for CO2 hydrogenation to hydrocarbons via FT

Used environment

The code was developed on Windows 10 but it should be platform independent.

  • Python version: 3.9.12 (amd64)
  • Packages:
    • torch:1.12.0
    • torchdyn:1.0.3
    • numpy:1.23.0
    • scipy:1.8.1
    • pandas:1.4.3
    • matplotlib:3.5.2
    • tqdm:4.64.0
    • doepy:0.0.1
    • notebook:6.5.3

Install the dependencies with pip install -r requirements.txt

Authors

Acknowledgement

Financial support from German Federal Ministry of Education and Research (BMBF) through the project InnoSyn (FKZ: 03SF0616B) and from German Research Foundation (DFG) through the project "NFDI4Cat - NFDI for Catalysis-Related Sciences" (DFG project no. 441926934) within the National Research Data Infrastructure (NFDI) programme of the Joint Science Conference (GWK) is gratefully acknowledged.