Charles Darwin University

CDU eSpace
Institutional Repository

 
CDU Staff and Student only
 

Machine vision for autonomous welding using three dimensional scanning

Lam, Thompson (2015). Machine vision for autonomous welding using three dimensional scanning. Bachelor of Engineering Co-op (4th Year Project) Thesis, Charles Darwin University.

Document type: Thesis
Citation counts: Google Scholar Search Google Scholar

Author Lam, Thompson
Title Machine vision for autonomous welding using three dimensional scanning
Institution Charles Darwin University
Publication Date 2015
Thesis Type Bachelor of Engineering Co-op (4th Year Project)
Subjects ENGINEERING
Abstract The use of robots for welding is beneficial as they can allow more rapid welding of parts and to also eliminate the harm exposed to workers arising from hazardous working conditions. These robots however are generally operated by using fixed sequences of commands that offer no flexibility to changes in the welding task. A review into the existing approaches to remedy this has provided an insight into the use of different sensor technology used to allow a welding robot to become more adaptable to variations in the tasks given to welding robots. The study identified the use of three dimensional (3D) scanners with 3D feature detection algorithms as a novel approach to allow identification of weld seams of simple parts without requiring any information about the weld surface geometry beforehand. Research made into various 3D scanning technology showed laser scanning as most suitable for this task due to its previously demonstrated ability to produce scans with high accuracy and resolution by other researchers. A prototype laser scanner developed from low cost parts confirmed the feasibility of scanning metal parts with high accuracy provided the surface was not overly reflective. 3D crease detection algorithms then applied to scanned models as well as computer generated models showed successful extraction of tee-joints from point cloud data. The crease detection was also shown to be reliable under random noise of up to 1mm applied in the x, y and z direction. These results demonstrate the feasibility of using 3D scanners to obtain weld seam locations allowing welding robots to weld objects of unknown geometry or where existing weld seam data is prone to deviations.
Keyword autonomous welding
laser scanning
3D point clouds
feature extraction
Additional Notes Full text restricted.
 
Versions
Version Filter Type
Access Statistics: 66 Abstract Views, 173 File Downloads  -  Detailed Statistics
Created: Tue, 29 Sep 2015, 08:51:00 CST by Jessie Ng