Journal of Additive Manufacturing Technologies
Vol 2 No 1 (2022): J AM Tech
https://doi.org/10.18416/JAMTECH.2212684

Articles, ID 684

Defect detection with image processing and deep learning in polymer powder bed additive manufacturing systems

Main Article Content

Ecesu Arslan (Arcelik A.S., Istanbul, Turkey), Deha Unal (Arcelik A.S., Istanbul, Turkey), O. Akgün (Turkish-German University, Istanbul, Turkey)

Abstract

Selective Laser Sintering (SLS) is a type of additive manufacturing process which uses a laser to fuse polymer particles on the powder bed. The process critically relies on controlling the heat and uncontrolled thermal gradients can cause the parts to curl during the process, which may fail the ongoing build with a cost. This layer-wise manufacturing process needs to be monitored during the build to ensure the process is free of problems. In this paper, deep learning-based defect detection system has been developed to detect any defect (curling, part shifting, short feed). The developed detection system aims to detect the existence of the anomaly during powder bed fusion process and send instant error notifications to the operator. Detection of the anomaly is a binary classification problem and it is solved with the scratch model. The novelty of the developed defect detection algorithm is, it can work independently from the size of the build area, the shape of the part and the location of the part. The developed model have training and test accuracy of, respectively 98.43% and 99.2%. Grad-CAM algorithm was used for the visual explanation of the model.  This study showed the detection system's effectiveness developed by the deep learning method without continuous human supervision in polymer powder bed additive manufacturing processes.

Article Details

How to Cite

Arslan, E., Unal, D., & Akgün, . O. (2023). Defect detection with image processing and deep learning in polymer powder bed additive manufacturing systems . Journal of Additive Manufacturing Technologies, 2(1), 684. https://doi.org/10.18416/JAMTECH.2212684