Automatised quality assessment in additive layer manufacturing using layer-by-layer surface measurements and deep learning

Le Roux, Léopold, Liu, Chao, Ji, Ze ORCID: https://orcid.org/0000-0002-8968-9902, Kerfriden, Pierre ORCID: https://orcid.org/0000-0002-7749-3996, Gage, Daniel, Feyer, Felix, Körner, Carolin and Bigot, Samuel ORCID: https://orcid.org/0000-0002-0789-4727 2021. Automatised quality assessment in additive layer manufacturing using layer-by-layer surface measurements and deep learning. Presented at: 14th CIRP Conference on Intelligent Computation in Manufacturing Engineering 2020, Naples, Italy, 15-17 July 2020. Procedia CIRP. , vol.99 Elsevier, pp. 342-347. 10.1016/j.procir.2021.03.050

PDF - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (976kB)

Abstract

Additive manufacturing (AM) has gained high research interests in the past but comes with some drawbacks, such as the difficulty to do in-situ quality monitoring. In this paper, deep learning is used on electron-optical images taken during the Electron Beam Melting (EBM) process to classify the quality of AM layers to achieve automatized quality assessment. A comparative study of several mainstream Convolutional Neural Networks to classify the images has been conducted. The classification accuracy is up to 95 %, which demonstrates the great potential to support in-process layer quality control of EBM.And the error analysis has shown that some human misclassification were correctly classified by the Convolutional Neural Networks.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Published Online
Status:	Published
Schools:	Professional Services > Advanced Research Computing @ Cardiff (ARCCA) Schools > Engineering
Publisher:	Elsevier
ISSN:	2212-8271
Date of First Compliant Deposit:	28 May 2021
Date of Acceptance:	31 May 2020
Last Modified:	01 Aug 2024 13:43
URI:	https://orca.cardiff.ac.uk/id/eprint/141624

Citation Data

Cited 5 times in Scopus. View in Scopus. Powered By Scopus® Data

Actions (repository staff only)

Edit Item