Enhancing Piping and Instrumentation Diagram Recognition: A Machine Learning Approach

This Study presents an effective method for automating the identification of symbols in Piping and Instrumentation Diagrams (P&ID). Utilizing the You Only Look Once (YOLOv8) object detection technique within the Roboflow platform, the study encompasses comprehensive data collection, annotation, and model training. The dataset, annotated using MakeSense AI, ensures precision in recognizing a diverse range of instrumentation symbols commonly found in process engineering. The resulting YOLOv8 model achieves high accuracy in symbol recognition. The integration of an HTML-based UI facilitates real-time identification of symbols in uploaded P&ID images, with an additional camera function for on-the-fly recognition of physical P&ID sheets. Emphasizing performance optimization through software-based solutions, the methodology demonstrates potential applications for advancing automation in the realm of process engineering.