MEP domain object classification through interdomain rule-based semantic enrichment on knowledge graphs

Abstract: The data and information of objects in building information modelling (BIM) from different software are delivered with incomplete data and misclassified objects. This study focuses on classifying mechanical, electrical, and plumbing (MEP) objects based on interdomain topological relationships and geometry conditions through semantic web technologies and semantic enrichment with a rule-based inferencing technique. Four rule sets were developed and run over 32 knowledge graphs of building models with at least 90% accuracy. False positives and negatives arose from non-discrete geometry and topology features of the objects. To address this issue, future work will integrate the proposed method with image recognition.