Automated Construction Specifications Digitalization for Efficient Compliance Checking on Sites

Abstract

To ensure construction quality and the safety of workers, construction managers oversee projects in accordance with established construction specifications. Manual review of the specifications requires considerable time and effort due to the dispersion of diverse criteria across extensive documents. Although various studies have aimed to digitalize construction documents for computerized analysis, they have focused on the design phase. Furthermore, compliance checking has only covered the predetermined rules, not over the entire specifications. This research expands the scope to encompass both qualitative and quantitative criteria automatically, enabling construction managers to access comprehensive specification contents digitally, for instance, through mobile devices. The research introduces a methodology for the automatic digitalization of specification items into appropriate formats including original texts, checklists, and Python codes. The methodology involves categorizing each specification item into one of four predefined types (i.e., general references, subjective judgment criteria, adaptive guidelines, and Pass/Fail standards). Subsequently, the methodology employs generative pre-trained transformers (GPT) to write checklists for subjective criteria, and Python codes for adaptive guidelines and Pass/Fail standards. The methodology ensures that the full set of construction specifications can be automatically converted into formats that allow for easy and accurate compliance checking by construction managers. It is expected to facilitate the straightforward and precise comprehension of complex construction standards across different activities through digitized construction specifications.