Heterogeneous network modeling and segmentation of building information modeling data for parallel triangulation and visualization

Abstract Building Information Modeling (BIM) is a promising technology for the construction industry, and BIM models have become widely accepted as 3D construction models. Currently, BIM data are often organized by product level to support cross-platform parallel and real-time visualization. However, cross-platform visualization of BIM products with large volumes of geometric data still poses a challenge to both triangulation and rendering processes. Existing efforts have mainly used geometric simplification and geometric data streaming technologies. This study addresses this issue from a different perspective by partitioning original large-scale BIM products into small BIM sub-products. First, a novel heterogeneous geometric relationship model (HeGeo) is proposed to categorize BIM relationships into reference, decomposition, and association relationships according to Industry Foundation Classes. On top of the HeGeo, a strategy for partitioning the geometric data of the original BIM products is proposed. Empirical studies were conducted on nine BIM models with a BIM product embedding all the geometric data. The experimental results showed that the proposed scheme improved triangulation efficiency 3.05 ± 0.57 times with the same hardware resources and the same triangulation tool through a parallel computing framework and raised the triangular mesh loading efficiency 1.53 ± 0.29 times by processing requests concurrently. In addition, the proposed scheme improved the user experience of online BIM visualization tools through incremental rendering. The segmentation scheme was generalized to product-level and floor-level schemes by constraining the segmentation nodes to BIM products and floors respectively. Thus, the segmentation scheme could be applied to any BIM model and facilitate BIM adoption by all stakeholders during a building's life cycle.