Identifying Unlabeled 3D Components in BIM Models for Industrial Projects

• Author / Creator

  Abdollahnejad, Sina

• Oil and gas projects are known for their size and complexity and incorporate multiple disciplines such as concrete, steel, and piping. Each discipline is executed within a confined area during a limited timeframe. The execution of each discipline requires careful planning and coordination between the different disciplines. Each discipline creates separate Building Information Models (BIM), which are merged together with others into one huge model in terms of complexity. This model is used for different purposes such as: coordinating work packages and detecting any possible clashes.
  From the contractor perspective, the BIM model can be utilized for defining the scope and obtaining a preliminary estimate while the project is in its early stages. The model’s value depends on its degree of completeness and time that it will be available. However, lack of standard structure for Building Information Modeling in the industry causes immature, inconsistent and incomplete BIM models during the early stages of the project. This means the model usefulness becomes limited, thus the contractor has to review the model manually to extract the required and useful information, including the scope of each discipline, a preliminary estimate of quantities, etc.
  The objective of this research is to investigate and develop a new methodology that can automatically fill the missing data in the BIM model and leverage its usage. This objective is achieved by identifying the type for each BIM model component using convolutional neural networks. This approach focuses on different projections of BIM model components rather than incomplete descriptive attributes. The research reviews different 3D image classification methods to select the most suitable method. After selecting a suitable method, an image classifier is developed to identify the missing labels for the BIM model components. Then, the methodology is validated by using three real-world industrial project models. Results indicate that the proposed method can automatically process ill-defined and incomplete BIM models to fill the missing data, and works with 91% accuracy on classifying the BIM model components.

• Subjects / Keywords

  • BIM
  • Convolutional Neural Networks
  • Image classification

• Graduation date

  Fall 2020

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-c1a5-rz07

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.