A Complete Piping Identification Solution for Piping and Instrumentation Diagrams

This study proposes a pipeline identification method for identifying pipelines in Piping and Instrumentation Diagrams (P&IDs) in image format. Automating this process is an important issue for the process plant industry, as image P &IDs are currently identified manually by humans. As part of research on technology for automatic conversion of image-format piping and instrumentation diagram (P&ID) into digital P&ID, this study proposes a method combining deep learning and traditional image detection to identify various types of pipelines in image format P&ID. The proposed method includes image pipeline extraction, pipeline simulation data augmentation, pipeline identification and fusion. For multiple test P &IDs, using yolov8 based on split coordinate spatial attention and augmented with simulated data, the average precision and recall are 90.5 % and 88.5 %, respectively. Combined with LSD line detection, it takes an average of 100 seconds to recognize a complete image, and the overall pipeline recognition rate can reach 95.5 %.