Robotizing Welding Manufacturing via Human-Robot Collaboration
and Deep Learning
Abstract
Skilled welders often outperform welding robots. To improve welding robots using human intelligence, the weld pool which welders respond to must be sensed. However, its specular nature disqualifies diffuse reflection-based laser triangulation methods. To overcome this issue, the mirror surface is advantageously used to reflect a laser pattern away from the arc, simultaneously eliminating the arc illumination problem. To allow welders to freely demonstrate their skills, a human-robot collaborative system has been established where a welder carries a virtual torch, similarly as operating an actual one, without a sensor. The movement is measured at the virtual system and then followed by robot which carries sensors and performs actual welding. The measured weld pool is displayed to the operator at the virtual site such that the welder can observe the change in the operation result to adjust his/her torch movement and other parameters. The true intelligence of the welder is thus contained for being extracted using deep learning. For more complex welding processes requiring multiple tools, their robotization is more challenging. A possible solution is also to learn from human welders as they are quicker learners who can adjust their operations to stabilize the complex welding process to generate training data.
Short Biography
Dr. Yu Ming Zhang is the James Boyd Professor of Electrical Engineering and College of Engineering’s Director of International Partnerships at the University of Kentucky (UK). His research focuses on robotizing welding processes through machine vision-based intelligence for intelligent robotic and human-robot collaborative welding systems. His research has brought him 12 US patents and over 200 journal publications. His recognition includes Fellow of AWS, ASME, SME, and IEEE and Dean’s Award for Excellence in Research from UK College of Engineering. Five of his graduate students won the IIW (International Welding Institute) Henry Granjon Prize on behalf of the US against IIW member countries’ national winners for dissertation/thesis research. Dr. Zhang is currently an Area Editor for the Journal of Manufacturing Processes published by the SME. He is, and has been, Associate Editor/Editorial Board Member for a number of major international journals including the IEEE Transactions on Automation Science and Engineering.