Model Validation Using Invariant Signatures and Logic-Based Inference for Automated Building Code Compliance Checking

Abstract

Fully automated building code compliance checking (ACC) requires accurate information extraction from both building information models (BIMs) and building code chapters, and equally (if not more) importantly, a precise matching between the two. Although research on information extraction has been extensively conducted for ACC, there is a lack of investigation of automated and practical information mapping between the extracted information, from BIMs to building code requirements. To address this gap, the authors proposed a new method for BIM model validation to validate an input Industry Foundation Classes (IFC) model with regard to building code concepts. This validation method was supported by creating invariant signatures of architecture, engineering, and construction (AEC) objects that capture the geometric nature of the objects. Target concepts from building codes are classified into four categories: (1) explicit concepts, (2) inferable concepts, (3) user-assisted concepts, and (4) system defaults. Identification algorithms are developed for all four categories based on the invariant signatures of AEC objects. An experiment was conducted to test the proposed method on validating five real commercial project models with selected concepts from a current building code. Compared with a manually developed gold standard, 99.8% precision and 99.6% recall were achieved. This demonstrates that the proposed method is promising in supporting information matching between BIMs and building code concepts for ACC purpose.