2024 AIChE Annual Meeting

(664a) Translating Process Flow Diagrams to Piping & Instrumentation Diagrams (P&IDs) with Artificial Intelligence

Authors

Schweidtmann, A. M. - Presenter, Delft University of Technology
Goldstein, D., Delft University of Technology
Schulze Balhorn, L., Delft University of Technology
Hirtreiter, E., Delft University of Technology
Piping and instrumentation diagrams (P&IDs) are important depictions of chemical and biochemical processes containing schematics of a process’ equipment, piping networks, and control structures. Developing P&IDs from process flow diagrams (PFDs) is a crucial step in process development, but typically involves a tedious, manual workflow.

We propose a methodology based on artificial intelligence for the prediction of control structures of P&IDs from PFDs, drawing inspiration from transformer-based models used in translating human language. This method treats the forecasting of control structures as a translation problem, transforming PFDs into P&IDs. Both flowsheets are represented as strings using the SFILES 2.0 notation. This makes them compatible with existing transformer-based translation models. Initially, our model undergoes pre-training on synthetic P&IDs to learn the syntax. Subsequently, it undergoes a finetuning step through transfer learning on industrial P&IDs.

References

[1] Hirtreiter, E., Schulze Balhorn, L., & Schweidtmann, A. M. (2024). Toward automatic generation of control structures for process flow diagrams with large language models. AIChE Journal, 70(1), e18259.