Anomalyspy: A Generative Defect Localization in Semiconductor Packages, with X-Ray Microscopy

Anomalyspy: A Generative Defect Localization in Semiconductor Packages, with X-Ray Microscopy

2025 IEEE 27th Electronics Packaging Technology Conference (EPTC)

10.1109/EPTC67330.2025.11392584

https://doi.org/10.1109/eptc67330.2025.11392584

Riko I Made, Pawan Kumar, Pablo Quijano Velasco, Ng Chee Koon, Leong Chang Jie

artificial intelligence, Machine Learning, failure analysis, semiconductor failure analysis, generative models, Defect Detection

24 February 2026

Made, R. I., Kumar, P., Velasco, P. Q., Koon, N. C., & Jie, L. C. (2025). Anomalyspy: A Generative Defect Localization in Semiconductor Packages, with X-Ray Microscopy. 2025 IEEE 27th Electronics Packaging Technology Conference (EPTC), 1–5. https://doi.org/10.1109/eptc67330.2025.11392584

Semiconductor packaging reliability is critical for advanced electronics, yet detecting rare defects like voids and misalignment remains challenging due to their subtlety and volumetric complexity. Traditional inspection methods often struggle with 3D X-ray microscopy (XRM) data, necessitating advanced computational approaches. This study introduces anomalyspy[1], a machine learning framework that leveraged Vector Quantized Variational Autoencoder (VQ-VAE) to detect anomalies in XRM images of semiconductor packages. By reshaping the z-dimension into channels (zkernel) and using 3D peak detection, we achieve precise defect localization. Trained solely on a minimally defective sample, our model demonstrates robust anomaly detection, offering a scalable solution for semiconductor failure analysis

Publisher Copyright

This research / project is supported by the A*STAR - Industry Alignment Fund - Pre-Positioning: Machine Learning Guided Failure Analysis & Diagnostic Capability Development for Next Generation 3D-IC Packaging (3D-IC)
Grant Reference no. : M22K8a0048, OUNI231001bENT-PP

This research / project is supported by the A*STAR - Advanced Manufacturing and Engineering (AME) Programmatic Fund: Mat-GDT (Materials - Generative Design Testing) Framework
Grant Reference no. : OUNI241001aENTMTC

This research / project is supported by the A*STAR - Towards Realistic Deep Learning for 3D Vision
Grant Reference no. : SC24/23-810600 EC-2023-046

This research / project is supported by the A*STAR - Industry Alignment Fund - Pre-Positioning: Singapore Battery Pack Programme (SGBP2) – WP2
Grant Reference no. : M23L6a0020, OUNI231001aENT-PP

This research / project is supported by the Energy Market Authority - Megawatt Hour Scale Sodium-ion Battery System for Solar Storage and Load Shifting
Grant Reference no. : ESME250101aPUBESS

This research / project is supported by the A*STAR - Manufacturing, Trade, and Connectivity Programmatic Fund
Grant Reference no. : M23L7b0021

© 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

https://oar.a-star.edu.sg/communities-collections/articles/22878

NA

Institute of Materials Research and Engineering