Semi-automatic representation of design code based on knowledge graph for automated compliance checking

Abstract

Automated compliance checking (ACC) intends to verify the compliance of designs in construction industry by design codes. The ability to interpret and represent semantic information of design codes determines the maximum application scope of ACC. However, design codes are clause texts written in natural languages and existing ACC studies usually use relatively low-complexity code clause samples. At present, the lack of an accurate representation model for design codes leads to difficulties in representing the implicit information, nested logic, and complex relations contained in high-complexity clauses in codes. To address this problem, this research establishes a new representation model based on knowledge graph (KG). Four schemas are proposed into the model including order, complex, event and integration schemas. Further, an accompanying methodology for semi-automatic construction of design code KG (DCKG) is proposed. It includes four parts: interpretation, reconstruction, organization, and implementation. Where the implementation part develops a code annotation platform. In the case study and experiment, a scenario of checking a building information model (BIM) of metro station by GB50157–2013 Code for Design of Metro is adopted to validate the newly proposed representation model and the automated compliance process. The results show that the proposed model and method are correct and feasible, and our model outperforms other models in the representation ability of design codes.