US 12,240,180 B2
Methods and systems for detection of impurities in additive manufacturing material
Alexander J. Coco, Chicago, IL (US); Brianna K. Nord, Chicago, IL (US); Robert W. Grube, Seattle, WA (US); Emma Romig, Chicago, IL (US); Aaron C. Drollette, Chicago, IL (US); and Eric M. Chapman, Chicago, IL (US)
Assigned to The Boeing Company, Chicago, IL (US)
Filed by The Boeing Company, Chicago, IL (US)
Filed on Feb. 28, 2020, as Appl. No. 16/804,629.
Prior Publication US 2021/0268740 A1, Sep. 2, 2021
Int. Cl. B29C 67/00 (2017.01); B29C 64/393 (2017.01); B33Y 10/00 (2015.01); B33Y 30/00 (2015.01); B33Y 50/02 (2015.01); B29C 64/153 (2017.01)
CPC B29C 64/393 (2017.08) [B33Y 10/00 (2014.12); B33Y 30/00 (2014.12); B33Y 50/02 (2014.12); B29C 64/153 (2017.08)] 20 Claims
OG exemplary drawing
 
1. A method for detection of impurities in additive manufacturing material during forming of a mechanical part, the method comprising:
illuminating, by a first light source and a second light source controlled by a computing device, a top layer of the additive manufacturing material with light during forming of the mechanical part;
while illuminating the top layer of the additive manufacturing material with light provided by the first light source at a first wavelength between 100 nanometers (nm) and 500 nm, causing a camera to acquire first image data of the top layer;
while illuminating the top layer with light provided by the second light source at a second wavelength outside the first wavelength, causing the camera to acquire second image data of the additive manufacturing material;
processing, using a learning model, the first image data and the second image data to determine an amount of impurities in the top layer, wherein the learning model is generated by supervised learning using training images based on manually identified impurities and uses edge detection, color extraction, and Laplacian operators to reliably distinguish impurities from additive manufacturing material;
generating, using a neural network, a pair of pixel-wise masks based on the first image data and the second image data, wherein individual pixels of at least one image of the first image data and at least one image of the second image data are assigned a debris label or a non-debris label;
based on the debris and/or non-debris labels represented by the pair of pixel-wise masks, calculating a percentage of the top layer that includes impurities;
performing, by the computing device, a comparison between the percentage of the top layer that includes impurities and a predefined threshold; and
based on the comparison, causing, by the computing device, one or more machine components to remove the top layer and trigger a new layer of different additive manufacturing material to be added to the mechanical part.