Building Information Modeling (BIM) integrates geometric, semantic, and relational information to support interdisciplinary collaboration. The Industry Foundation Classes (IFC) standard ensure interoperable data exchange; however, practical BIM models often contain incomplete semantics and ambiguous relationships. These limitations restrict automated rule-checking and reduce the transparency required for reproduci-ble computational workflows, particularly when spatial and topological relationships are implicitly represented (Nepal et al., 2012). This thesis proposes a graph-based framework that reformulates BIM rule-checking as a pattern-matching and graph-transformation process. IFC models are converted into a property graph, where building elements are represented as nodes and rela-tionships as edges; enabling explicit and traversable structures for reasoning. Based on this representation, modular rule engines evaluate semantic, connectivity, and ge-ometric conditions. The framework is implemented as a pipeline comprising graph transformation, en-richment, rule evaluation, and structured output generation. The evaluation through rule-based scenarios demonstrates robustness under incomplete IFC data. The re-sults show that graph-based representations establish transparent and extensible BIM validation, providing a foundation for explainable, intelligent BIM systems built on re-usable decision patterns.     «

Building Information Modeling (BIM) integrates geometric, semantic, and relational information to support interdisciplinary collaboration. The Industry Foundation Classes (IFC) standard ensure interoperable data exchange; however, practical BIM models often contain incomplete semantics and ambiguous relationships. These limitations restrict automated rule-checking and reduce the transparency required for reproduci-ble computational workflows, particularly when spatial and topological relationsh...     »