US 12,325,191 B2
Additive manufacturing with in situ defect detection
Steven M. Storck, Catonsville, MD (US); Nathan G. Drenkow, Columbia, MD (US); Brendan P. Croom, Baltimore, MD (US); Ryan H. Carter, Ellicott City, MD (US); and Robert K. Mueller, Columbia, MD (US)
Assigned to The Johns Hopkins University, Baltimore, MD (US)
Filed by The Johns Hopkins University, Baltimore, MD (US)
Filed on Feb. 21, 2022, as Appl. No. 17/676,407.
Claims priority of provisional application 63/151,656, filed on Feb. 20, 2021.
Prior Publication US 2022/0266531 A1, Aug. 25, 2022
Int. Cl. B33Y 50/02 (2015.01); B29C 64/153 (2017.01); B29C 64/393 (2017.01); B33Y 10/00 (2015.01); G06T 7/00 (2017.01); G06V 10/70 (2022.01)
CPC B29C 64/393 (2017.08) [B29C 64/153 (2017.08); B33Y 10/00 (2014.12); B33Y 50/02 (2014.12); G06T 7/0004 (2013.01); G06V 10/70 (2022.01); B29C 2791/009 (2013.01); G06T 2207/10048 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/30144 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for monitoring and analyzing an additive manufacturing process, the method comprising:
heating, via an energy source, a melt zone to form a melt pool to fuse an additive media on an active layer to build a part being manufactured based on a part design model;
receiving, by a sensor, reflected radiation from the melt pool due to the energy source acting upon the melt pool and generating raw melt data of the melt pool based on the reflected radiation;
generating, based on the raw melt data, an active layer dataset that is spatially defined;
analyzing the active layer dataset with respect to a plurality of defect signatures within a defect signature library to identify matches between the active layer dataset and the plurality of defect signatures, the defect signatures within the defect signature library indicating characteristics of a melt pool that are indicative of the formation of a defect due to conditions including temperatures below a target temperature for the melt pool and temperatures above the target temperature for the melt pool, the defect signature library being predefined based on a machine learning processing of historical sensor datasets with corresponding ground truth datasets; and
detecting a defect in the part based on the analyzing the active layer dataset with respect to a plurality of defect signatures.