1. Data-driven Digital Twin for Board-Level Packaging Interconnects under Multi-physics Loading
- Author
-
Jing Luo, Yang Liu, Ke Li, Zhen Pan, Chiyuan Ma, and Jicun Lu
- Abstract
Solder joints of electronic packaging devices are used as mechanical fixation and electrical interconnection between chips and circuit boards, which provides protection for the normal operation of electronic equipment. Therefore, real-time monitoring of the status of solder joints is essential for predictive maintenance of equipment. In this paper, we propose a digital twin based on semi-supervised learning for diagnosing faults in chip interconnection solder joints. In order to achieve maximum generalization of limited label information, the interdependence between sample labels with similar feature distributions is fully exploited by semi-supervised learning. Additionally, we use real-time monitoring data to update the learning model, and reveal the evolution of solder joint failure under different loads through diagnostic results of model. A dynamic model is formed by stimulated fault evolution characteristics into a solder joint failure model to form a diagnose fault dynamic model in a virtual space. Finally, we designed a thermal-vibration coupling experiment to verify the effectiveness of the digital twin-based solder joint failure diagnosis model. The results show that the digital twin maintains good consistency with the performance degradation process of the solder joint throughout its life cycle. Moreover, the diagnostic accuracy of the digital twin model can reach 85%, which proves that our method can monitor the service status of physical entities online, and intelligently predict the failure mode and life cycles under load conditions.
- Published
- 2023