This repo contains the code used in the paper Solvent pre-selection for extractive distillation using the Gibbs-Helmholtz Graph Neural Network submitted to ESCAPE-33. This work uses the version v2.0.0 of GH-GNN.
To cite this work use:
@incollection{sanchez2023solvent,
title={Solvent pre-selection for extractive distillation using Gibbs-Helmholtz Graph Neural Networks},
author={Sanchez Medina, Edgar Ivan and Sundmacher, Kai},
booktitle={Computer Aided Chemical Engineering},
volume={52},
pages={2037--2042},
year={2023},
publisher={Elsevier}
}
To cite GH-GNN use:
@article{sanchez2022ghgnn,
title={Gibbs-Helmholtz Graph Neural Network: capturing the temperature dependency of activity coefficients at infinite dilution},
author={Sanchez Medina, Edgar Ivan and Linke, Steffen and Stoll, Martin and Sundmacher, Kai},
journal={Digital Discovery},
DOI={10.1039/D2DD00142J},
year={2023},
volume={2},
issue={3},
pages={781-798},
publisher={RSC},
}
Solvent preselection using the GH-GNN model can be easily carried out using the solvent_preselection class contained in the file solvent_preselection.py.
As described in the above mentioned paper, the selction can be performed by either:
- Relative volatility at infinite dilution
- Minimum solvent-to-feed ratio
Both methods are conviniently implemented as functions of the class solvent_preselection.
Case studies for the separation of aliphatic/aromatic and olefin/paraffin mixtures are provided on the other respective files.
from solvent_preselection import solvent_preselection
mixture = {
'c_i': {'smiles':'CCCCCC', 'name':'n-hexane'},
'c_j': {'smiles':'c1ccccc1', 'name':'benzene'},
'mixture_type': 'aliphatic_aromatic',
'T_range': (25 + 273.15, 85 + 273.15),
}
solvents = [...] # list of solvents SMILES
AD = ... # Applicability domain strategy to be applied to GH-GNN (either 'both', 'class', 'tanimoto or None)
sp = solvent_preselection(mixture, solvents, AD)
sp.screen_with_rv()
sp.screen_with_minSF()
You could also give a star ⭐ to this repo!
Note that this code has an MIT license that needs to be respected at all times