BIM 부재 및 IFC 클래스 간 시멘틱 매핑 검증을 위한 기계학습기반 건축 부재 자동 분류

With the growing use of BIM in the construction industry, new software applications are being developed to meet these specific needs. Such developments have increased the importance of the Industry Foundation Classes (IFC), which is the open and neutral format for sharing BIM data and the de facto standard for interoperability. However, the IFC lacks formal logic rigidness and thus is susceptible to errors and omissions during data transfers. This research addressed the issue by applying machine learning techniques to automatically classify elements of a BIM model, thereby allowing for classification checks as well as enriching the semantics of the model. Naïve Bayes, logistic regression and support vector machines (SVM) were trained and tested using 4,187 unique elements from six BIM models. Features included elements’ geometries and the relational semantics between elements. Results showed that SVM provided the highest accuracy at 0.9081 with geometry, and 0.9439 when adding semantics. The algorithms provide a way to accelerate integrity checks and model element classification required for quality control of BIM models.