Artur NOVODRANOV, Ukraine (online)

Applying neural network technologies for quality control in robotic additive manufacturing

3D printing technology is currently actively used to manufacture parts in various industries. The production of metal parts using additive deposition may be accompanied by the appearance of surface defects in the deposited layers, namely pores and cracks. Using a robotic complex based on an anthropomorphic robot with an integrated machine vision system increases the surfacing process’s quality and productivity. The hardware component of the machine vision system consists of a high-speed GigE camera and a controller. The software uses the YOLO-NAS neural network architecture to detect and classify defects automatically after the completion of the deposition process of the corresponding layer. The quantitative characteristics of the quality of recognition and classification using a trained neural network model are given. It is recommended that anthropomorphic robots and machine vision systems based on neural network technologies be used for manufacturing parts for a responsible purpose using WAAM technology.

Leave a Comment

Event Details
  • Start Date
    08.10.2024 16:10
  • End Date
    08.10.2024 16:30
  • Status
    Expired
  • Category