Status: Expect regular updates and bug fixes.

OpenAI Gym environment for solvent extraction process design and simulation

Learning environment for solvent extraction process design.

Links

• Source code repository: https://github.com/solventx/gym-solventx

Installation

You can install the package directly from GitHub with following commands:

git clone https://github.com/solventx/gym-solventx.git
cd gym-solventx
pip install -e .

Dependencies

Cantera, Gym, SciPy, Numpy

Using the module

The module can be imported as a normal python module:

import gym_solventx

Try out the demo here.

Solvent extraction process design leaderboard

Successful design: Recovery > 0.8, Purity > 0.985

Name	Process configuration	Episodes before solve	Purity,Recovery (mean,std dev)	Agent details	Train script
Siby Jose Plathottam	Input: Nd, Pr, Ce, La	2000	Purity: 0.99,0.005 Recovery: 0.85,0.04	PPO with RNN

Issues

Please feel free to raise an issue for bugs or feature requests.

Who is responsible?

Core developer:

• Siby Jose Plathottam splathottam@anl.gov
• Nwike Iloeje ciloeje@anl.gov

Contributor:

• Blake Richey blake.e.richey@gmail.com

Acknowledgement

Citation

If you use this code please cite it as:

@misc{gym-solventx,
  title = {{gym-solventx}: Gym Environment containing a solvent extraction process simulator.},
  author = "{Siby Jose Plathottam, Blake Richey, Nwike Iloeje, }",
  howpublished = {\url{https://github.com/solventx/gym-solventx}},
  url = "https://github.com/solventx/gym-solventx",
  year = 2019,
  note = "[Online; accessed 21-August-2019]"
}

Copyright and License

Copyright © 2019, UChicago Argonne, LLC

Process Design Environment Simulator for Reinforcement Learning (SolventX) is distributed under the terms of BSD-3 OSS License.