Semantic BIM enrichment using a hybrid ML and rule-based framework for automated tenement compliance checking

Abstract

Semantic enrichment enhances BIM models by extracting structured information, improving their applicability for Automated Code Compliance Checking. Rule-based methods rely on well-defined conditions but struggle with tasks like space classification, where explicit checking rules are unavailable. Meanwhile, ML-based classification introduces adaptability but faces liability challenges due to misclassifications. This paper proposes a hybrid framework integrating ML for space classification and rule-based inferencing for tenement identification. The approach ensures that ML automates preprocessing, improving classification through meticulous feature engineering, while rule-based reasoning guarantees logical consistency during verification. Validated using real-world datasets from residential projects in Mumbai, India as a case, the ML-based space classification component achieves an F1-score of 0.85 and accuracy of 0.86, demonstrating its effectiveness. The deterministic tenement identification process delivers error-free results for various dwelling configurations, making it highly suitable for verification workflows. This study advances scalable BIM-based compliance systems by refining semantic enrichment methodologies for future applications.