A Self-Guided Mold Assembly System Based on Deep Learning for Engineering Training

• Published in: 2022 International Conference on Engineering Education and Information Technology (EEIT) pp. 84 - 87
• Main Authors: Pan, Xudong, Huo, Hong, Cai, Lvyin, Lv, Jianfeng, Han, Qianghui, Pan, Yuan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2022
• Subjects: Deep learning; engineering training; Information technology; mold assembly; Production; Real-time systems; Streaming media; Tracking; Training
• DOI: 10.1109/EEIT56566.2022.00027

Smart assembly is a part of smart manufacturing that, with the rapid development of the latter, has become a weak link in modern production chains. The corresponding training experiments for mechanical assembly and disassembly used in colleges and universities need to integrate new technologies to improve the quality of education being imparted to engineering students. This study develops a self-guided system for mold assembly that is suitable for engineering training. It can identify the parts of a given system in images and project them to increase our knowledge of them. A deep learning network model is designed to facilitate the replacement of the assembly that recorded an accuracy of identification above 99% in experiments. A hand detection and tracking model is built in Python that can detect and track the real time movements of hands in a video at a speed of recognition of up to 37 FPS. The proposed system is used in an engineering training course to verify its accuracy of recognition of assemblies. It improves the students' interest in learning owing to its novel teaching methods and helps cultivate their ability for autonomous learning.