Doctor of Philosophy
Faculty of Engineering
Nicholson, Alexander, Rapid adaptive programming using image data, PhD thesis, Faculty of Engineering, University of Wollongong, 2005. http://ro.uow.edu.au/theses/422
This work was initiated to address the problem of automated weld repair in a specific application; namely hydro turbine cavitation damage. An extensive review of the application requirements was undertaken and this highlighted the need for a new approach to robot programming. This thesis therefore focused on the development of a novel rapid programming technique based on a vision system. The implementation involved the development of algorithms to correlate robot and vision system coordinates, communication links to the system hardware, and a software environment to ‘manage’ the process. The complete system was successfully validated by representative trials on a full-size hydro turbine mock-up. Several investigations examining different industrial applications such as the weld repair of hydro turbine runners in the power generation industry and weld repair of crawler shoes in the mining sector have demonstrated the feasibility of using robot technology. However, limitations have been identified in these investigations that continue to impede the widespread acceptance and use of this technology in industry. An offline programming technique was developed in this thesis that addresses some of these limitations, and seeks to improve confidence in robotic systems for ‘one-off’ or small batch run applications. OH&S regulations were found to be the significant driver in developing systems for insitu repair by robot. Small commercial robots are available that partially satisfy the requirements for in-situ use; however, a fast means of robot programming and the ability to handle uniquely damaged shapes at arbitrary locations remains unsolved. These have been appropriately addressed in the current work with the development of an automatic adaptive system capable of rapid robot programming and online process control via an intuitive, user-friendly PC-based interface. Various sensor technologies and profilometry tools and techniques were reviewed. Based on simplicity, cost, and availability, the most promising approaches identified were touch-sensing or non-laser-based vision for profilometry, followed by offline robot programming based on the acquired data. In this thesis, image acquisition was achieved using a low-cost CCD camera, damaged areas were identified, and weld paths automatically generated. A combined and integrated profilometry and offline programming technique was devised in conjunction with a robotic system incorporating touch-sensing. The speed of the system developed in this thesis enables positional repair welds of unique shape to commence within minutes, making the prospect of ‘one-off’ repairs commercially viable. The technique developed is suitable for the target application; hydro turbine cavitation repair, but can be easily extended to many other wear replacement scenarios commonly found in industry.