|M.Sc Student||Pevzner Alexei|
|Subject||Automated 3D Point Cloud Sensing and Real-Time Dynamic|
Projection for Operation Monitoring and Quality
|Department||Department of Autonomous Systems and Robotics||Supervisor||ASSOCIATE PROF. Amir Degani|
Lean Construction (LC) and Building Information Modeling (BIM) are two rapidly growing applied research areas that strongly impact construction control. LC aims to maximize the value and eliminate the wastes in the construction process, while BIM aims for greater collaboration among the teams of a project during its design and construction phases.
While construction techniques and materials have advanced, humans still bear much of the workload executed on site. Work methods still include time-consuming and error-prone tasks like monitoring, progress evaluation and error detection. These tasks are carried out mostly manually by highly paid professionals. Yet at the same time, numerous Construction Tech (CT) innovations have been proposed. Measuring devices, such as high-precision laser scanners, can be applied to reduce cycle time, waste, construction errors and the rework that necessarily follows.
The main objective of this research was to demonstrate that efficient data flow from site to BIM and vise-versa leads to significant improvement of the construction process. An application of scanning technology is proposed to provide workers with real-time feedback regarding the quality and accuracy of their handwork, with the goal of achieving proper quality the first time, with fully automated inspection, and no rework. If needed, further iterations can be executed, as the flow is cyclic. The termination condition depends on the operator's quality demands. As a proof of concept, the system is demonstrated using an example of wall plastering. The result of the plaster application is difficult to measure, and errors are difficult to detect. Plaster application on walls is prone to surface errors like flatness, where the plane is not flat within specified tolerances, and perpendicularity, where the wall is not perpendicular within specified tolerances to another wall. Such errors are difficult to detect with the naked eye. The proposed system monitors the progress of the procedure, evaluating the surface flatness and projecting corrections onto the surface itself, after optimizing with respect to industry standards and tolerances. The proposed method is demonstrated on an experimental setup, in a rectangular room using a Trimble TX8 laser scanner capturing a 3D point cloud and an adjustable projector. Using RANSAC and K-means algorithms, the main planes of the room are detected and validated using BIM information. For a chosen wall, optimizing over two degrees of freedom, an optimal wall plane is calculated to minimize the total cost of wall mending, namely, shaving-off protrusions and filling up depressions. Next, a realistic topographic map is produced with respect to the optimal plane. The map is then projected pre-warped onto the wall, properly displaying the working personnel which areas should be mended. The results show high precision detection of wall flatness deviations, of up to 2 mm, which is lower than the permitted tolerances of the British Standards Institution.
In this thesis, the technology setup is described, the experiments conducted, and the results. There is also a discussion of the broader concept and the ways in which it can be applied by either manual or robotic work methods.