Bhat, DarshankumarDarshankumarBhatMünch, StefanStefanMünchRöllig, MikeMikeRöllig2024-05-162024-05-162024https://publica.fraunhofer.de/handle/publica/46793510.1109/EuroSimE60745.2024.10491479In today’s fast-paced and highly interconnected world, digitization is a significant driving force across various sectors, with electronic systems being backbone of this. Increasing number of electronic devices are being manufactured which introduces significant reliability challenges. Prognostic and Health Monitoring (PHM) is one such discipline which employs methods to develop a strategy of electronic systems monitoring and failure prediction. It is a challenging task to establish such framework which combines the precursors, information on failure mechanisms, and quick prediction which allows users to be aware of the impending failures in advance. In this article, we propose a tailored and tested PHM approach for failure prediction of solder joints, one of the most prone elements in electronics. A machine learning model is first trained with the synthetically generated solder damage data. It is then utilized in the PHM framework, whose primary focus lies on the realization of life estimation of solder joints as a portable solution. The methodology is implemented with two microcontrollers including comprehensive study on model conversion, model compression and model degradation. The framework is then validated with the field mission profile data acquired from the eBikes and associated data pre-processing steps and memory limitations are discussed.enMicromechanical devicesMicrocontrollersMachine LearningLife estimationMicroelectronicsReliabilitySolderingconference paper