edfas.org 33 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 27 NO. 1 A SUMMARY OF THE ISTFA 2024 PANEL DISCUSSION: AI APPLICATIONS IN FAILURE ANALYSIS Ted Kolasa* and Greg Johnson** ISTFA 2024 Panel Discussion Organizers *Northrup Grumman, Chandler, Ariz. — ted.kolasa@ngc.com **Carl Zeiss Microscopy, White Plains, N.Y. — greg.johnson@zeiss.com This year’s Panel Discussion was held on Halloween afternoon to a packed house at the San Diego Bayfront Hotel. In keeping with the theme of the conference, the chosen topic was Artificial Intelligence in Electronic Failure Analysis. An esteemed panel of four experts was on hand to field questions from the audience: Sebastian Brand of Fraunhofer IMWS, Michael Campbell, from Qualcomm CDMA Technologies, Zachary Eich of AMD, and Flavio Cognigni of Zeiss Microscopy. The Panel Discussion began with brief opening statements from each of the panelists. Sebastian Brand expressed that AI can improve the efficiency of failure analysis (FA) by extracting valuable insights from reports, images, and signals and in this way can help preserve valuable FA expertise. AI can assist in tasks such as acoustic waveform analysis and image classification, freeing the analyst from the processes of extracting critical details from qualitative and quantitative data, which can often be tedious or challenging for humans in other ways. This makes it a valuable tool, aiding experts in critical decisionmaking during failure analysis. Michael Campbell highlighted the data handling challenges in the semiconductor industry, where vast amounts of data are generated. Connected engineering databases can streamline data extraction and improve decisionmaking. He envisions a future where AI-based systems help automate FA by providing context-specific insights from vast datasets and minimizing knowledge gaps. Campbell also highlighted the potential to refine past learnings through continued optimization of AI algorithms. Zach Eich shared two examples of AI applications in FA. The first involved using computer vision to detect and classify solder ball failures, such as cracks or pad cratering, with high precision. The second case used AI to detect thermal interface material voids in images, helping track failures to train and improve FA systems. The 2024 Panel Discussion featured four panelists and a lively Q&A session.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==