This paper proposes using transformer-based language models to automatically predict P&ID control structures from process flow diagrams, casting the problem as a translation task using the SFILES 2.0 notation. The approach recognizes reusable patterns in existing P&IDs and leverages NLP architectures for process engineering. P&IDs are foundational documents for MOC because they define the physical configuration of process plants. Any change to equipment, piping, or control systems must be reflected in updated P&IDs and assessed for safety implications. AI-assisted P&ID generation and consistency checking can ensure that proposed changes maintain design integrity, automatically flag deviations from engineering standards, and accelerate the technical documentation updates required as part of MOC closure.
