A text mining based automated BIM model data checking system

Building Information Modelling (BIM) has been widely used in the construction industry describing geometry of the building, spatial relationships and geographic information, as well as the quantities and properties of BIM objects and their properties. With large amounts of data exported from building information models, BIM data checking usually costs large amounts of man-hours to guarantee BIM model quality. In order to improve accuracy and efficiency for model data checking, applying machine learning techniques is highly expected. The machine learning algorithms can be used by combining data from many buildings, the characteristics, and location of flats to automatically detect flaws of building design, or even the likelihood of construction delays. Anomaly detection is useful to pinpoint modelling errors. In this project, semantic natural language processing technologies embedded with a machine leaning engine is developed to realize (1) automated object and property information retrieve from the massive BIM mode data, (2) similar objects recognition and naming consistence checking, and (3) automated BIM model data completeness and accuracy checking. The machine learning engine is wrapped by a dashboard-based user interface to facilitate manual object/property checking or batch data checking. The specially designed modules of the machine learning engine are allowed to be customized to support different construction regulatory requirements.