Automatic Classification and Coding of Prefabricated Components Using IFC and the Random Forest Algorithm

Abstract

The management of prefabricated component staging and turnover lacks the effective integration of informatization and complexity, as relevant information is stored in the heterogeneous systems of various stakeholders. BIM and its underlying data schema, IFC, provide for information collaboration and sharing. In this paper, an automatic classification and coding system for prefabricated building, based on BIM technology and Random Forest, is developed so as to enable the unique representation of components. The proposed approach starts with classifying and coding information regarding the overall design of the components. With the classification criteria, the required attributes of the components are extracted, and the process of attribute extraction is illustrated in detail using wall components as an example. The Random Forest model is then employed for IFC building component classification training and testing, which includes the selection of the datasets, the construction of CART, and the voting of the component classification results. The experiment results illustrate that the approach can automate the uniform and unique coding of each component on a Python basis, while also reducing the workload of designers. Finally, based on the IFC physical file, an extended implementation process for component encoding information is designed to achieve information integrity for prefabricated component descriptions. Additionally, in the subsequent research, it can be further combined with Internet-of-Things technology to achieve the real-time collection of construction process information and the real-time control of building components.

Keywords

• prefabrication
• IFC
• Random Forest
• automatic coding
• information classification