Deep learning based visual automated sorting system for remanufacturing

Nwankpa, Chigozie E. and Eze, Solomon C. and Ijomah, Winifred I.; (2020) Deep learning based visual automated sorting system for remanufacturing. In: 2020 IEEE Green Technologies Conference (GreenTech). IEEE, USA, pp. 196-198. ISBN 9781728150178 (

[thumbnail of Nwankpa-etal-IEEE-2020-Deep-learning-based-visual-automated-sorting-system]
Text. Filename: Nwankpa_etal_IEEE_2020_Deep_learning_based_visual_automated_sorting_system.pdf
Accepted Author Manuscript

Download (173kB)| Preview


Remanufacturing is a crucial component of the circular economy concept which emphasises sustainable consumption habits. This study proposes a novel automated sorting system for remanufacturing which is based on deep convolutional neural networks(CNN). To demonstrate its applicability, the proposed deep learning (DL) system was used to distinguish among dry, wet, oily and defected surfaces. The test was conducted on four locally sourced 3" x 6" plates. Sample image data were captured using a USB webcam. The network training was done with 75% of the data while the balance data were used for testing. In this preliminary study, the DCNN classified the features with up to 99.74% accuracy on validation data and above 96% accuracy on live video feed; demonstrating that it can accurately sort components. This study is the first to propose a low-cost sorting system for remanufacturing based on the deep CNN and logic gates. The results show that the method is an accurate, reliable, cost-effective and fast technique that can potentially outperform existing sorting systems in the remanufacturing industry.