Web-Books
im Austria-Forum
Austria-Forum
Web-Books
Informatik
Joint Austrian Computer Vision and Robotics Workshop 2020
Seite - 122 -
  • Benutzer
  • Version
    • Vollversion
    • Textversion
  • Sprache
    • Deutsch
    • English - Englisch

Seite - 122 - in Joint Austrian Computer Vision and Robotics Workshop 2020

Bild der Seite - 122 -

Bild der Seite - 122 - in Joint Austrian Computer Vision and Robotics Workshop 2020

Text der Seite - 122 -

PowderBedAnalysis inAdditiveManufacturingUsingImageProcessing FlorianRecla,MartinWelk UMIT–PrivateUniversity forHealthSciences,Medical InformaticsandTechnology florian.recla@edu.umit.at, martin.welk@umit.at Abstract. Systems for additive manufacturing are experiencing an enormous upswing in the industry. In thispaperamethodfor theopticalcontrolofpow- derbeds ispresented. Thesystemisbasedonacam- eraanddirectionallightingandissuitablefordetect- ingtwotypesofdefects, including(i)areaswheretoo little/toomuchpowderhasbeenapplied, and (ii) ar- eas with different porosity. The system is evaluated forboth typesof errors. 1. Introduction Binder-Jetting is a popular method for additive manufacturing of high-resolution components. In this process, powder is applied in layers, which is then selectively cured by a binder [3]. In order to prevent dead times and production downtimes in powder-bed-based additive manufacturing, a system was sought that would reliably find defects in the in- dividual powder layers. Defects in the powder bed can occur either in the form of excess/missing pow- der, or as insufficient porosity of the powder. If such defects are not detected, components may be pro- duced which do not achieve the expected strength values or contain predetermined breaking points in- side. Theanalysis systemshouldbesimple indesign and reliable in operation. Three different types of optical analysis are used in existing plants: Laser triangulation [2], a cam- era with structured illumination [4] or a camera with directed illumination [1]. All approaches aim at creating a geometric image of the powder surface. Laser triangulation or structured illumination can be usedtocreate three-dimensionalmodelsof theplane, while directed light can only be used to find qual- itative deviations from the plane. Since a qualita- tiveevaluationof thesurfaceissufficient, themethod withdirected light is chosen. Top-light Side-light Shadows Figure1. Side-light causes shadowsatdefectivepositions 2. ImagingSystem The prototypical image system is installed in an existingmachine for additivemanufacturing. The system consists of a camera and two lights. One light source illuminates the powder bed verti- cally fromabove toachieve themostuniformillumi- nation possible (top-light). The second light source shines on the powder bed at a very flat angle (side- light). Theside-lightcreates shadowswhen thereare differences in height in the powder bed, which the camera captures from above. These shadows are not created when using the top-light (Figure 1). After the creation of each powder layer, two images are acquired, one using the top-light, and one using the side-light. By subtracting the two images from each other, the shadows are extracted and evaluated. The top-lightensures thatcolordifferences in thepowder arenotmisinterpretedas shadows. 3. ImageProcessing After acquiring the images of one layer, the pro- cessing of the images is done. The two images arehigh-passfilteredtominimiseglobal illumination differences. (Aconstantcorrectionisnotpossiblebe- cause differences in the powder mixture lead to dif- ferent reflection properties.) Afterwards the images are subtracted from each other to extract the shad- ows. A powder layer without defects thus produces an image with very low grey values, as there is no difference between top-light and side-light. Defects 122
zurück zum  Buch Joint Austrian Computer Vision and Robotics Workshop 2020"
Joint Austrian Computer Vision and Robotics Workshop 2020
Titel
Joint Austrian Computer Vision and Robotics Workshop 2020
Herausgeber
Graz University of Technology
Ort
Graz
Datum
2020
Sprache
englisch
Lizenz
CC BY 4.0
ISBN
978-3-85125-752-6
Abmessungen
21.0 x 29.7 cm
Seiten
188
Kategorien
Informatik
Technik
Web-Books
Bibliothek
Datenschutz
Impressum
Austria-Forum
Austria-Forum
Web-Books
Joint Austrian Computer Vision and Robotics Workshop 2020