Case Study: transformer-based solution for the automatic digitization of gas plant

The paper presents a solution based on Generative Artificial Intelligence, developed by Engineering to automate the digitalization of SNAM’s gas transportation facilities by extracting and organizing information from technical documentation.

The solution integrates Vision LLMs, Object Detection, and Transformer models to analyze engineering documents and Piping and Instrumentation Diagrams (P&IDs), identify plant components, and reconstruct their hierarchical structure.

Experimental results demonstrate 91% accuracy in textual data extraction, 93% accuracy in component identification, and approximately 80% accuracy in hierarchical structure reconstruction, highlighting the potential of AI to automate industrial processes and accelerate the digitalization of energy infrastructure.