Semantic Building Information Modeling: An empirical evaluation of existing tools

Abstract

Semantic Building Information Modeling (BIM) consists in translating data expressed using BIM formats (namely IFC) into Semantic Web files using RDF serializations (e.g., Turtle). This enables the inference of new knowledge and constraint checking, among other advantages. While several software tools for translating BIM models into Semantic Web languages have been proposed in the literature, they differ in the features exposed.

This paper analyzes and empirically compares some of these tools (namely, IFC converters translating an input IFC model into an RDF graph), identifying their strengths and main limitations. Our methodology includes measuring computation times of common tasks (file conversion, query and inference over output files), assessing the retention of knowledge (particularly, geometric information) and examining reasoning capabilities (complexity and completeness of the resulting models). Our results show that IFCtoLBD is the best option in many cases. IFCtoRDF and IFC2LD are slower but better preserve geometric information, while KGG is faster at the expense of losing information in the translation.