A Hybrid Prognostics and Health Management Model to Predict Drop-induced Degradation in Ball Grid Array Interconnects

(23:20 + Q&A) Chris Martinez, ME Dept, U-Maryland — Physics based health degradation models have been used for decades across industry and research. Reliability Physics models, such as the Coffin Manson Relation for cyclic fatigue, utilize an understanding of the stress-state and the resulting material degradation. These physics-based models utilize stochastically significant amounts of experimental data with inherent variability across specimens, and account for them in the models with population statistics (e.g. mean, standard deviation, kurtosis, etc.). While such models are well suited for design of large population of products, they are of limited value when it comes to personalized degradation monitoring and prognostics of individual pieces in that population. Such individualized Prognostics and Health Management (PHM) tasks benefit from machine learning (ML) approaches that allow personalized real-time model fitting (and updating) based on the degradation behavior of each specimen. However, an underlying understanding of the degradation physics offers opportunities for improving the confidence bounds of the PHM ML model. In this study the PHM approach is demonstrated for electronic printed circuit assemblies (PCAs), containing Ball Grid Array (BGA) components soldered to printed circuit boards (PCBs), experiencing progressive cumulative fatigue degradation under repeated drop/shock loading. The drop excitation and resulting histories of response deformation and incremental electrical degradation are monitored. Based on this data, a ML model is trained to predict degradation in the BGA interconnects, using feature engineering and a foundational understanding of the underlying degradation physics. When tested against verification datasets, the ML model is seen to provide satisfactory degree of accuracy.
Bio: Chris A. Martinez is a mechanical engineering M.S. student at the University of Maryland, College Park in the Center for Advanced Life Cycle Engineering (CALCE). Chris received his B.S. from the University of Maryland as well in 2024. Chris’s previous experience and research interests focused mainly in manufacturing. He worked at the Eaton Corporation in their aerospace machined components division as a manufacturing engineering co-op. Afterward, he worked for Prof. Davis J. McGregor’s MIRAGE Lab at University of Maryland and contributed to machine learning projects for quality control. He is currently responsible for reliability characterization of printed hybrid electronics at extreme conditions under his advisor Dr. Abhijit Dasgupta. He aspires to incorporate AI/ML more into the electronics reliability space as well as integrate AI/ML into large scale manufacturing operations.

Edited videos and slides from most of the REPP talks are available at https://attend.ieee.org/repp/?page_id=2110
Join our Dlist to hear about future REPPs: https://attend.ieee.org/repp/?page_id=361
Organized by the IEEE Silicon Valley chapter of the Electronics Packaging Society: https://ieee.org/scveps