An augmented MapReduce framework for Building Information Modeling applications

Lifecycle and effective dada management has become and still remains a difficult task in architecture, engineering, and construction domain. Recently, the emerging Building Information Modeling (BIM) gains more and more momentum in construction data presenting, processing, and sharing. With precise 3D digital geometry graphs, BIM has the potential to integrate diverse data from construction phases (design, construction, maintenance) and therefore becomes a promising collaboration platform for involved personnel. High efficiency BIM model processing is one of the critical issues for construction industry to truly embrace BIM technology. However, this topic has been rarely studied compared with other issues in BIM research. This paper presents an augmented version of MapReduce (MR), the popular distributed parallel computing model, to solve this problem. The main contribution lies in two points: 1) BIM models are extracted into uneven partitions according to construction domain features, followed by a specifically designed pre-process mechanism, which solves the unsuitability of classical MR´s even blocking in construction domain and further greatly improves performance; 2) process and thread level parallel computing techniques are introduced into MR in single node to form a two-tier hybrid parallel architecture, which is more adaptive to BIM´s massive graphical data processing. The proposed framework has been successfully used in real business applications of project quantity computation and BIM model collision detection. Experiment result from lab environment proves its efficiency and usability.