Dynamic Multi-Dimensional BIM for Total Construction As-Built Documentation

Jaehyun Park, Purdue University

Abstract

This dissertation first describes the design of a new, multi-dimensional (nD) building information modeling (BIM) structure to facilitate the as-built documentation. A work breakdown structure (WBS)-based database design is devised to enable the creation of a dynamic nD BIM database to incorporate construction records in a timely manner. The construction records are defined as an additional dimension, and a WBS code-based automated linking mechanism between construction project tasks and BIM objects is created to generate the multi-dimensional BIM database. The automatic linking mechanism and the WBS-based dynamic nD BIM database are tested and verified through the implementation of a real construction project schedule and a BIM model. Next, the dissertation describes the visualization of the nD BIM that contains the timely construction records. In the proposed nD BIM database structure, the relationship between the schedule and the BIM (4D BIM) is concentrated as the visualization of the nD BIM because the visualizations of the relationship can vary as time passes. An alternative Web- and database-supported visualization method is created, which enables real-time information sharing through a daily 4D BIM model in the nD BIM database. A central database structure that contains an automatic status determination and update method for each BIM object visualization is designed to facilitate the display of daily 4D BIM. The newly created method is demonstrated through the implementation in a real construction project, using its schedule and BIM model. The third part of the dissertation describes new methods created to timely incorporate construction records in the nD BIM. Among various types of construction records, two types—photos and daily reports—are selected to develop the method for construction record insertions. They are two main categories of construction records. During the construction phase, daily reports and photos are recorded on a daily basis and contain useful information that is critical to project success. First of the three tasks in this part, an automated daily photo registration method to the 4D BIM model in the nD BIM database is proposed as the method for inserting photos into the nD BIM database. A large number of as-constructed photos are taken during the construction phase for documentary purposes. Traditionally, the contents and context of the photos are manually grasped—a practice that is inefficient and is prone to errors—and results are not shared among project participants in time. A promising alternative is to register the photos to the 4D BIM model in the nD BIM database to identify BIM objects in the photos and retrieve their contents in the digital form. In the method, a content-based image retrieval (CBIR) approach is devised to find the closest photographed location by comparing photos with the 4D BIM model, and then a 2D grid concept is used to extract BIM objects from the model that is associated with the photos. The method is validated through the implementation of photos of a real construction project. Second of the three tasks, an automated daily photo contents and context retrieval method using the 4D BIM model in the nD BIM database is proposed to utilize the registered photos in the nD BIM database. The proposed retrieval method is developed for detecting problematic objects and categorizing the problems among the identified BIM objects that are related to the photos. Comparing photos to the images from the 4D BIM model using CBIR analysis leads to accurate identification of the problematic objects in the photos. In the proposed method, a neural network-based image style transfer algorithm is employed to reflect the 4D BIM image styles to the photos. The Gabor and Tamura texture features are applied to execute texture feature-based CBIR analysis, and the Chi-square similarity concept is introduced to measure similarities between the photos and the images from the 4D BIM model. A novel process for identifying problematic objects in photos is also developed. The method is validated using as-constructed photos of a real construction project. Third of the three tasks, an automated method for updating construction progress information to the 4D BIM model in the nD BIM database through a Web environment is developed to record and utilize daily reports in the nD BIM database structure. In the current practice, updating tasks of the 4D BIM model are accomplished manually. A promising alternative to the manual method is to update the 4D BIM models using Web-based daily reports that contain daily construction progress information. Based on the proposed central database in the second part of the dissertation, an extended 4D BIM database, an automatic model update method, and a Web interface have been developed. The newly developed method is tested through cases from a real construction project. (Abstract shortened by ProQuest.)

Degree

Ph.D.

Advisors

Cai, Purdue University.

Subject Area

Architectural engineering|Information Technology

COinS