Prognostic and health management of die-delamination in QFN packages
Over the past several decades, advancements in microelectronics technology have given rise to integrated circuits (ICs). External factors such as temperature changes and thermal shocks can eventually undermine the IC reliability leading to malfunctions and failures.
To cope with the costs associated with the lack of reliability, the electronic industry is increasingly adopting prognostic and health management (PHM) techniques. These techniques generally rely on physics-based models or data-driven ones. Unfortunately, both have limitations. Indeed, physics-based models are often incomplete since they cannot model all aspects which influence real-world degradation processes whereas data-driven models may not be a viable option when run-to-failure data is scarce.
To address this shortcoming, within the iRel40 project Sirris and Imec decided to explore the potential of hybrid techniques to fuse the strengths of both physics-based and data-driven methods.
More specifically, we propose a new framework based on a hybrid model PHM of an electronic system using as a case study a QFN package (see figure Figure 1) subject to die-delamination. To the best of the author's knowledge, our use case is the first to validate a hybrid approach for PHM at the device level.
Figure 1 Cross-sectional view of the QFN package
Our results show that the hybrid model outperforms models based only on physics or data-driven models. In this sense, our framework is suitable for the early detection of quality-relevant events such as die delamination. Furthermore, the framework can provide accurate predictions of the remaining useful life which can support predictive maintenance.
Alessandro Murgia (Sirris), Chaitra Harsha (Sirris), Elena Tsiporkova (Sirris), Bart Vandevelde (Imec), Chinmay Nawghane (Imec)
Ninix, On Semi, Plastic Omnium
Prognostic and Health Management, Reliability, Remaining Useful Life, QFN, die-delamination