Profile
Nationality : Chinese
Keywords : Image Processing, Welding technology, Deep Learning, Pulsed GMA welding
Research advisor : Prof. Satoshi Yamane
Why I entered GSC program
Time flies! I have been studying in Saitama University for nearly one year. I have benefited a lot from this program as follows. Firstly, Japan is a developed country with advanced science and technology, which have succeeded in addressing energy shortages. My major is about welding and deep learning. Welding is an important manufacturing technology and an energy-intensive process. High-quality welding technology and automation can help people avoid risks and save labor. More importantly, automated welding systems can improve the accuracy and efficiency of the welding process and reduce unnecessary energy consumption. This way is beneficial to the environment and I would like to do devotion. Secondly, the International Graduate Program on Green and Sustainable Chemical Technologies (GSC) covers a wide range of research fields, facilitating intercultural exchange among young individuals studying science. As a developed country, Japan has a lot of advanced technology and ideas to learn from. That is what I am seeking for, which inspires me to study further. Thirdly, GSC reduces students’ financial burden and allows us to better devote our attention to studies and research. We need not worry about our tuition and cost of living. In addition, I had a dream when I was a child. I was interested in Japanese culture and customs. I desired to study in Japan. In my opinion, studying abroad can broaden my horizons and lay a good foundation for my future work. I am grateful to have had such a great opportunity to realize my dream. I will cherish this time to make efforts to focus on my studies and research.
Research title
Application of Deep Learning to Welding Gap Identification in GMA Welding
Research abstract
In pulse Gas Metal Arc (GMA) welding, it is crucial to adjust welding conditions to accommodate changes in the root gap of the groove, ensuring high welding quality. The combination of visual sensors and deep learning technologies proves beneficial in estimating the gap. In this research, captured images of the molten pool across various root gaps were used. The identification of the gap was accomplished by using improved Resnet50. Specifically, the gap was successfully identified in groove welding with an 8mm gap, achieving high accuracy rate. Moreover, under the groove welding with the variation gap from 4mm to 8mm, this gap identification method also demonstrates excellent identification performance and satisfies real-time identification. The validity of the deep learning in GMA welding was verified.
Graphical abstract
