Australian Researchers Refine Robotic Blasting
Researchers in Australia recently published new work on the topic of robotic abrasive-blasting systems, developing an approach to automatic self-diagnosis of problems that could lead to greater ease of use and more widespread adoption of the potentially game-changing surface-preparation technology.
Andrew Wing Keung To, Gavin Paul and Dikai Liu, working out of the Centre for Autonomous Systems at the University of Technology in Sydney, published the paper, “A Comprehensive Approach to Real-Time Fault Diagnosis During Automatic Grit-Blasting Operation by Autonomous Industrial Robots,” in the February 2018 edition of Robotics and Computer-Integrated Manufacturing.
The Centre for Autonomous Systems has for more than a decade played a role in the development of autonomous blast systems, in 2013 creating the spinoff company Sabre Autonomous Solutions, dedicated to creating new technology for autonomous blasting on steel structures such as bridges.
The research team working on the new study looked at how a robotic blasting system might be equipped to self-diagnose problems in the field. Issues like hose blowouts, running out of abrasive media or a failure in the compressed-air supply can wreak havoc when a robotic blast system is at work, because unlike a manned system wherein the blaster can stop the job immediately to address the problem, an autonomous system might not stop working, resulting in lost time and a substrate that’s not evenly blasted.
The study they conducted looked at three approaches to detecting faults in the robotic blast system: RGB-D cameras (depth imaging), audio detection and pressure transducers.
RGB-D imaging uses a visual indicator to show where blasting is occurring and if a change can be detected in the process, such as if blast media is no longer coming out of the nozzle and therefore is not cleaning the substrate.
Sound, the researchers explain, can be used to automatically detect changes in the blasting process, such as if the system stops sending air and media through the nozzle, or if air continues but media stops.
Pressure detection can’t be used to differentiate between situations in which air and media are being pushed through the hose, but can be used to tell if air has stopped and blasting has ceased for whatever reason.
Each approach has some limits—audio detection, for example, can be less effective in situations where other blasting is occurring nearby or other sources of sound are present.
All three methods of detecting faults, though, proved more than 95 percent effective, leading the team to concluding that further research on combining the three systems could yield an effective method for automatically detecting failures in a robotic blasting system.
The entire paper is available to read for free here.