An Artificial Intelligence Based Approach for Instrument Index Generation from Piping and Instrumentation Diagrams

Summary This work tackles a significant challenge in energy sector engineering, procurement, and construction (EPC) workflows, where piping and instrumentation diagrams (P&IDs) are often shared in paper or portable document format (PDF) without detailed instrument information. The objective is to develop a comprehensive artificial intelligence (AI)-driven tool that automatically generates instrument indices from P&IDs, enabling quick identification and understanding of the instruments represented. The method involves instrument detection using fine-tuned AI models for text detection, specifically the progressive scale expansion network (PSENet) deep learning text detector and computer vision techniques, and text recognition via a fine-tuned PaddleOCR (optical character recognition) framework. Developed using over a dozen real-world projects’ P&IDs, the application has shown high accuracy, exceeding 99% for studied formats, significantly reducing manual effort. By digitizing and structuring instrument information, the solution streamlines project execution, supports rapid decision-making, and lays the groundwork for scalable AI applications in engineering documentation. It offers a practical, low-cost approach to P&ID digitization, with direct benefits for instrumentation planning, procurement, and maintenance in energy industry projects.