Open Access

Application of machine vision for the detection of powder bed defects in additive manufacturing processes

Marcin Korzeniowski

Marcin Korzeniowski

Department of Metal Forming, Welding and Metrology, Faculty of Mechanical Engineering, Wroclaw University of Science and TechnologyWroclaw, Poland
Search for this author on
Sciendo|Google Scholar
Korzeniowski, Marcin
, 
Aleksandra Małachowska

Aleksandra Małachowska

Department of Metal Forming, Welding and Metrology, Faculty of Mechanical Engineering, Wroclaw University of Science and TechnologyWroclaw, Poland
Search for this author on
Sciendo|Google Scholar
Małachowska, Aleksandra
 and 
Marta Wiatrzyk

Marta Wiatrzyk

  • Department of Metal Forming, Welding and Metrology, Faculty of Mechanical Engineering, Wroclaw University of Science and TechnologyWroclaw, Poland
  • Scanway LTDWrocław, Poland
Search for this author on
Sciendo|Google Scholar
Wiatrzyk, Marta
  
Sep 04, 2023
Application of machine vision for the detection of powder bed defects in additive manufacturing processes's Cover Image
About this article
Previous Article
Next Article
Cite
Download Cover
Published Online: Sep 04, 2023
Page range: 214 - 226
Received: Apr 07, 2023
Accepted: Jun 02, 2023
DOI: https://doi.org/10.2478/msp-2023-0013
Keywords
© 2023 Marcin Korzeniowski et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Fig. 1.

View of the working platform with a scraper in the form of a blade: A) side view, B) top view)
View of the working platform with a scraper in the form of a blade: A) side view, B) top view)

Fig. 2.

A) Installed BVS002C camera with BAE000K illuminator, B) Camera above the working field
A) Installed BVS002C camera with BAE000K illuminator, B) Camera above the working field

Fig. 3.

A) Software configuration of the captured image and restriction of the analysis area to a circle and B) automatic recognition of the area with extracted features
A) Software configuration of the captured image and restriction of the analysis area to a circle and B) automatic recognition of the area with extracted features

Fig. 4.

Views of distributed powder using different processing variables
Views of distributed powder using different processing variables

Fig. 5.

Application of the segmentation algorithm to extract features corresponding to irregularities in the distributed powder at different scraper travel speeds: A) Input image, 40 mms/s, B) Segmented image, 40mm/s, C) Removal of the load from the scraper, input image, 100 mms/s, D) Segmented image, 100 mm/s
Application of the segmentation algorithm to extract features corresponding to irregularities in the distributed powder at different scraper travel speeds: A) Input image, 40 mms/s, B) Segmented image, 40mm/s, C) Removal of the load from the scraper, input image, 100 mms/s, D) Segmented image, 100 mm/s

Fig. 6.

Image processing algorithm used in the developed application
Image processing algorithm used in the developed application

Fig. 7.

View of the powder layer distributed in an unacceptable way and classified by algorithm as NOK. A) Input image with limited region of interest, B) output image
View of the powder layer distributed in an unacceptable way and classified by algorithm as NOK. A) Input image with limited region of interest, B) output image

Fig. 8.

View of the powder layer speeded acceptably and classified by algorithm as OK. A) Input image with limited region of interest, and B) output image
View of the powder layer speeded acceptably and classified by algorithm as OK. A) Input image with limited region of interest, and B) output image

Fig. 9.

Image classification algorithm used in the study
Image classification algorithm used in the study

Fig. 10.

A),B) View of the powder layer distributed in an unacceptable manner with the anomaly maps and C) powder distributed acceptably
A),B) View of the powder layer distributed in an unacceptable manner with the anomaly maps and C) powder distributed acceptably

Example results obtained during the algorithm, with a view of the input images of the powder layer, the processed images, and the result of the classification

Input image Processed image Decision Comment
NOK many defects
OK few defects
NOK very many defects
Language:
English
Publication timeframe:
4 times per year
Journal Subjects:
Materials Sciences, Materials Sciences, other, Nanomaterials, Functional and Smart Materials, Materials Characterization and Properties
Journal RSS Feed