equdist

Welcome to the equdist repository! This project contains the code and resources necessary for implementing a Python-integrated JAX-written rigorous Newton-Raphson algorithm for equilibrium separation, with applications to distillation sequencing for purifying multi-component mixtures.

Table of Contents

• Introduction
• Features
• Installation
• Usage
• Project Structure
• Contributing
• License

Introduction

This project aims to provide an efficient and reliable method for distillation sequencing using a Newton-Raphson algorithm implemented in JAX. The primary application is to leverage reinforcement learning to optimize distillation processes, providing a scalable and flexible solution for chemical engineering tasks.

Features

• Rigorous Newton-Raphson Algorithm: Implemented in JAX for fast and reliable equilibrium separation.
• Automatic Differentiation: Supports automatic differentiation for efficient Jacobian calculation.
• Scalability: Tested with up to ten hydrocarbon components exhibiting ideal behavior, with potential for expansion to more complex mixtures.
• Thermodynamics: Leverages the DIPPR thermodynamic system and is compatible with the Aspen Plus thermodynamic database, making it adaptable to various chemical processes.

Installation

To run this project, the repository can be installed via;

pip install git+https://github.com/NiklasSlager/equdist.git

Usage

1. Run the Training Notebook: A Jupyter notebook named tutorial.ipynb is provided to run the model. A link to Google Colab is provide for easy testing and use.

2. Distillation input Parameters: The input parameters to the distillation column are:

   1. Number of stages
   2. Feed location
   3. Feed rate (kmol/hr)
   4. Feed composition
   5. Reflux ratio
   6. Distillate rate
   7. Operating pressure

3. Run the Algorithm: The plots visualize profiles including liquid composition, liquid and vapor flowrate, temperature, and liquid and vapor enthalpies.

Project Structure

The algorithm entails two steps for better convergence performance: In the first step, the equilibrium separation is solved assuming equimolar overflow. The heat equation is replaced with a total flow constraint to solve the Newton-Raphson procedure assuming equimolar overflow. The full set of MESH equations is solved in the second step.

• equdist/: Main directory containing the core code for the Newton-Raphson algorithm.
  • equimolar: Module for the equimolar overflow procedure.
  • model: Module for the full NR procedure.
  • functions auxilary functions used within the algorithm.
  • thermodynamics Description of the thermodynamic model using the DIPPR equations
  • costing Module for a basic total annualized cost estimation procedure based on the Marshall & Swift correlations
• notebooks/: Jupyter notebooks for example usage.
• README.md: Project documentation.

Contributing

Contributions are welcome! If you have suggestions for improvements or new features, please open an issue or submit a pull request. For major changes, please discuss them with the repository owner first.

License

This project is licensed under the MIT License. See the LICENSE file for more details.