Learn how to automate defect detection in serial production using supervised learning on BSE image datasets to identify features with high precision and recall.
Artificial intelligence (AI) techniques, including machine learning, are increasingly integrated into additive manufacturing workflows to interpret process data and analyze large datasets, such as layer-by-layer powder bed images. In powder bed fusion (PBF), each layer can be imaged and analyzed, providing a basis for data-driven assessments of process stability and part quality. In situ backscattered electron (BSE) imaging serves as an effective inspection technique, utilizing backscattered electron contrast from the solidified layer to detect density variations, cracking, and surface features.
By combining machine learning with electron beam-based imaging, this approach demonstrates the potential for scalable, in-situ defect detection in additive manufacturing. It establishes the groundwork for future remelting strategies capable of using melt pool depth to close previously detected defects above a certain threshold in the next layer.
Agenda:
Overview of how a supervised learning approach enables automated defect detection using sequential BSE image datasets, applicable for defect detection in serial production.
Discussion of the model training process on a manually annotated dataset, allowing the model to associate BSE data with defect likelihood.
Explanation of the inference process, where the trained detector evaluates new layer images, locates regions of interest, and outputs bounding boxes with associated confidence scores to assess defect likelihood.
Assessment of model performance through precision and recall metrics, ensuring reliable identification of actual defects while minimizing false positives.
Jonathan Buckley AM Applications Engineer, JEOL USA
Jonathan Buckley is an AM applications engineer for JEOL USA, supporting JEOL's additive manufacturing division. He has worked in the additive manufacturing industry for the past 10 years, primarily supporting electron beam powder bed fusion technologies. Buckley has experience both from supporting an EB-PBF machine manufacturer and from utilizing EB-PBF technology at an AM contract manufacturer focused on the production of additively manufactured orthopedic implants. He actively engages with the AM community through technical outreach including webinars, trade shows and industry events to educate peers on the advantages and applications of EB-PBF.
Comments
Post a Comment