Anzeige
Erschienen in:
2022 | OriginalPaper | Buchkapitel
Accelerated Defective Product Inspection on the Edge Using Deep Learning
verfasst von : Dimitris Tsiktsiris, Theodora Sanida, Argyrios Sideris, Minas Dasygenis
Erschienen in: Recent Advances in Manufacturing Modelling and Optimization
Verlag: Springer Nature Singapore
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
A common problem in casting products is the defects and the unwanted irregularities that may occur in the manufacturing process. Quality control is an important step, as it ensures that clients receive products free from such defects. It is often a time-consuming yet important task that detains the production chain. As the technology evolves, machine vision and deep learning algorithms can provide a significant benefit to industries. In this paper, we present a hardware accelerated deep learning approach for detecting irregularities on casting products via visual inspection. Several optimizations have been made to achieve high-end performance on an embedded scale. A proof of concept, that we specifically designed for industrial use, achieves 99.72% classification accuracy, is able to inspect up to 136 units per second and draws about 28 W of power on peak. The proposed embedded system can be seamlessly integrated into the assembly lines of factories.