(161b) Automated Synthesis and Generation of Optimal Process Flowsheets and Data Sheets with Enhanced Sustainability: Transformer and Reinforcement Learning Approaches

Automating the design of chemical processes is an important area of research, with the potential to significantly reduce the time and effort required for manual input of process data, while improving the accuracy and sustainability of process design. In this study, we investigate the use of two different approaches for automating the synthesis and generation of optimal process flowsheets and data sheets: Transformer models and reinforcement learning.

Our approach focuses on using SFILES software to optimize chemical process design, while the Transformer model interprets natural language input and converts it into machine-readable format for the software. We also investigate the potential for using reinforcement learning to further improve the accuracy of the generated process designs.

We explore the potential benefits of using a combination of the Transformer and reinforcement learning approaches, and discuss the limitations and challenges associated with each approach. Overall, our study highlights the potential of automation technology to revolutionize the field of chemical process design, and suggests that the use of Transformer models and reinforcement learning could be a promising avenue for future research and development.