<div> <div> <p>Building Information Modeling (BIM) serves as an important media in supporting automation in the architecture, engineering, and construction (AEC) domain. However, with its fast development by different software companies in different applications, data exchange became labor-intensive, costly, and error-prone, which is known as the problem of interoperability. Industry foundation classes (IFC) are widely accepted to be the future of BIM in solving the challenge of BIM interoperability. However, there are practical limitations of the IFC standards, e.g., IFC’s flexibility creates space for misuses of IFC entities. This incorrect semantic information of an object can cause severe problems to downstream uses. To address this problem, the author proposed to use the concept of invariant signatures, which are a new set of features that capture the essence of an AEC object. Based on invariant signatures, the author proposed a rule-based method and a machine learning method for BIM-based AEC object classification, which can be used to detect potential misuses automatically. Detailed categories for beams were tested to have error-free performance. The best performing algorithm developed by the methods achieved 99.6% precision and 99.6% recall in the general building object classification. To promote automation and further improve the interoperability of BIM tasks, the author adopted invariant signature-based object classification in quantity takeoff (QTO), structural analysis, and model validation for automated building code compliance checking (ACC). Automation in such BIM tasks was enabled with high accuracy.</p><p><br></p><p><br></p> </div> </div>