The intelligent development of oil and gas pipelines is crucial for overcoming the challenges of transitioning from informatization to intelligence, especially considering the vast number of standards governing pipeline operations. With over 3000 distinct standards collected, their multi-source heterogeneity makes effective manual use virtually impossible, leading to a significant semantic gap and difficulty in scenario-based retrieval. This study introduces a novel knowledge graph (KG) specifically designed to address this challenge by providing a unified, efficient method for managing and applying these standards. The KG is constructed based on a predefined ontology to provide an efficient semantic framework for technical specifications. We employ a novel hybrid information extraction framework, combining advanced neural models (BERT/BiLSTM-Attention) with rule-based methods and a Graph Convolutional Network (GCN) for relationship extraction. This approach resolves terminology ambiguity inherent in multi-source documents and accurately extracts entities and relationships from heterogeneous standards. This framework facilitates the construction of a comprehensive knowledge graph for pipeline network technical specifications, thereby enhancing the accessibility and usability of critical technical knowledge. The constructed KG, covering 17,655 entity nodes and 20,454 relationships, facilitates the intelligent retrieval of information and promotes the digital transformation of pipeline management. The validation demonstrates a high F1-Score in information extraction and confirms the capability of the KG to shift the knowledge retrieval paradigm from slow document searching to rapid, non-linear relational querying. By integrating data from diverse standards, this approach supports technological advancements, enhances safety and sustainability in the oil and gas industry.
Hui et al. (Sun,) studied this question.