About this Abstract |
Meeting |
2021 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2021)
|
Symposium
|
Poster Session
|
Presentation Title |
Anomaly Segmentation in Thermal Tomography Process Monitoring for Laser Powder Bed Fusion Using Texture Analysis |
Author(s) |
Alexander Groeger, Tanvi Banerjee, Joy Gockel, John Middendorf, Tom Spears |
On-Site Speaker (Planned) |
Alexander Groeger |
Abstract Scope |
Process monitoring in laser powder bed fusion automated by artificial intelligence will significantly help manufacturers save time by detecting processing anomalies and suggesting corrective actions when intolerable defects are identified. For these systems to perform analysis and draw conclusions, in-situ sensor imagery needs to be analyzed and segmented for anomalous activities within each layer. Image segmentation is performed on in-situ thermal tomography collected during an additive manufacturing build using texture analysis. This analysis uses a deep learning convolutional neural network to encode images into a low dimensional space describing combinations of textural features. This texture-feature space is visualized using dimensionality reduction to find groups that describe unique textural signatures. These groups can be labeled by field experts and compared to material characterization data, enabling the model to classify where thermal events are occurring on the build layer and identify defect locations in the fabricated material. |
Proceedings Inclusion? |
Definite: Post-meeting proceedings |