A method for clustering unlabeled BIM objects using entropy and TF-IDF with RDF encoding

Oil and gas projects involve different construction disciplines such as mechanical, structural, and electrical. The current practice in these projects involves creating separate building information models for the different disciplines and compiling them into one model to check for collisions or conflicts. Due to intellectual property, contractual requirements, unfinished design, or technical issues during final model compilation, the final merged model lacks essential data for contractors such as the trade of each object in the model. Nonetheless, the model is issued to contractors who utilize it in different pre-construction planning tasks. However, due to data loss, incompleteness, or inconsistency, the model usability can become limited and the contractor has to review the model manually to extract information from it. This is a lengthy and costly task that becomes more challenging in fast-tracked projects that involve periodically issuing updated Building Information Models. One type of information that contractors need for different planning and estimation purposes is the scope of work for different construction trades in different areas of the project. In many cases, models lack explicit attributes of 3D objects that make it possible to perform an automated query of these objects by trade type. This research suggests a state of the art solution to automate the extraction of this information in such cases. In this paper, we describe a method that utilizes Resource Description Framework (RDF) encoding of BIM data together with Term Frequency-Inverse Document Frequency (TF-IDF) and entropy-based algorithms to automatically group 3D objects based on their trade. The proposed methodology is tested using three actual cases of oil and gas projects with more than four million objects in total. The results show that the proposed method can achieve a 91% purity in the generated groups.