# A Reinforcement Learning Approach with Masked Agents for Chemical Process Flowsheet Design
This repository contains the code for the two case studies presented in the paper *A reinforcement learning approach with masked agents for chemical process flowsheet design*.
The work focuses on the generation, design and optimization of chemical process flowsheets using Reinforcement Learning. The full paper can be found in <https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.18584?af=R>
The work focuses on the generation, design and optimization of chemical process flowsheets using Reinforcement Learning.
The full paper can be found in <https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.18584?af=R>
## Case Study 1
This case study compares the performance of discrete and hybrid masked PPO agents in generating a chemical process flowsheet for the reaction $A \rightarrow B$. With this illustrative example it was found that for simple examples in which the number of discrete and continuous varibles are reduced, a fully discretized agent outperfroms the hybrid agent, achieving better rewards.
