edfas.org 35 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 27 NO. 1 ISTFA 2024 USER GROUP HIGHLIGHTS Daminda Dahanayaka,* Joy Liao,** and Anita Madan,*** ISTFA 2024 User Group Chair/Co-Chairs *IBM, Albany, N.Y. — daminda@ibm.com **Nvidia, Santa Clara, Calif. — jliao@nvidia.com ***Independent Consultant, New York, N.Y. — amadan16@gmail.com The ISTFA 2024 conference celebrated its 50th anniversary at the Hilton San Diego Bayfront from October 28 to November 1, 2024. A highlight of this landmark event was a series of six dynamic in-person User Group sessions. These sessions were carefully curated to promote active engagement, foster meaningful discussions, encourage networking, and create lasting opportunities for dialogue on key industry topics beyond the conference itself. Some sessions featured one or two compelling mini presentations aimed at initiating discussions on targeted topics, complemented by interactive exchanges between panelists and the audience. Other sessions focused on exploring current industry challenges and best practices in their respective areas. Participation was enthusiastic throughout all six sessions, with attendees—ranging from analysts at leading semiconductor companies to tool vendors and academic representatives—contributing insights and sharing experiences. The sessions were praised for their valuable content, interactive format, and high levels of engagement. Attendees were encouraged to continue these discussions through ASM community boards, supporting an ongoing collaborative exchange of ideas. ISTFA 2024 ARTIFICIAL INTELLIGENCE (AI) IN FAILURE ANALYSIS USER GROUP Chair/Co-Chairs: Peter Hoffrogge, Thomas Rodgers, Konstantin Schekotihin, Sebastian Brand, and Florian Felux peter.hoffrogge@pvatepla.com, thomas.rodgers@zeiss.com, konstantin.schekotihin@aau.at, sebastian.brand@imws.fraunhofer.de, florian.felux@infineon.com The primary objective of the second meeting of the AI in FA User Group at ISTFA was to stimulate curiosity and engage participants with emerging topics in AI. This year’s focus centered on showcasing select AI applications in signal and image analysis, as well as discussing the lessons learned by the chairs during the implementation of AI technologies in failure analysis laboratories. A key emphasis of the last part of the discussion was on the standardization of various aspects and components of IT systems and making them ready for AI-based applications. To motivate the audience and to start the discussion, the group decided to start with live demonstrations of the most prominent AI use cases, the application of convolutional neural networks to signal and image processing. These networks allow for the direct analysis of raw data obtained by FA methods and automate various failure identification workflows. In the first demo shown by Peter Hoffrogge, PVA TePla, the audience could see how image annotations created in the CVAT tool can be mapped onto raw waveforms recorded by a scanning acoustic microscope. The second demonstration, given by Thomas Rodgers and Sreenivas Bhattiprolu from Carl Zeiss Microscopy, provided an overview of the arivis Cloud and AI toolkit developed at Zeiss. The demonstration started with a short introduction to image segmentation and novel 3D defect detection. The example presented to the audience lays an important perspective for the AI applications in FA based on large foundational models that can easily be fine-tuned to identify lab-specific defects with only a few training examples. Thus, the foundational model presented at the demo session was able to learn the correct segmentation mask for images showing specific defects from just 10 to 15 annotated samples. In addition, fine-tuning the foundational model requires much fewer computational
RkJQdWJsaXNoZXIy MTYyMzk3NQ==