Advanced Search   
  Look in
       Title     Description
       Inactive Projects

Quality Assessment of As-built Building Information Models using Deviation Analysis
Head: Burcu Akinci and Daniel Huber
Contact: Engin Anil
Mailing address:
Carnegie Mellon University
Robotics Institute
5000 Forbes Ave
Pittsburgh, PA 15213
Associated lab(s) / group(s):
 3D Vision and Intelligent Systems Group
Quality assessment (QA) of as-built building information models (BIMs) is traditionally conducted by physically measuring key elements in the environment and comparing those measurements to the corresponding virtual measurements in the BIM. This research is investigating an alternative approach – using patterns in the differences between the data within and between steps in the as-built BIM creation process to identify potential errors.

In the architecture, engineering, and construction (AEC) industry, laser scanners are increasingly used to capture the geometry of buildings and infrastructure, since these devices generate dense, accurate 3D point measurements of the visible surfaces. In current practice, a client, such as an architect or engineering firm, will contract with a 3D modeling service provider to create as-built CAD models or BIMs from these raw “point cloud” measurements. The manual process involves several steps – data collection, data alignment, and 3D modeling. Errors can arise in each stage, and the purpose of QA is to detect significant errors before the final model is accepted by the client.

The physical measurement method for QA compares randomly sampled physical measurements to their BIM counterparts. This approach works well, but has a number of disadvantages, including the need for physical access to the facility, sparse coverage of potential error locations, and lack of intuition as to the source of errors.

This research is investigating an alternative approach, in which patterns in the differences between raw points and the final model or differences between points in scans taken from different locations are used to identify errors in the modeling process. For example, modeling errors are apparent when visualizing the difference in surface shape between the raw and modeled data. The proposed approach has the advantages of complete coverage, no need for physical access, and insight into the source and cause of the error. We are analyzing this approach and formalizing it for use within the General Service Administration’s (GSA’s) 3D-4D BIM program.

Sponsor: This work is supported by the GSA Grant #GS-00P-CY-P-0321.