Supervised machine learning models for accurate prediction of CBGA solder joint lifespan

This is the authors' abstract. We don't add key points for this paper.

This study introduces the first comparative analysis of multiple machine learning models for predicting CBGA solder joint lifespan under thermal cycling, a key factor in electronic reliability. Expanding the dataset from 45 to 295 reveals that data augmentation significantly improves prediction accuracy. Basically, DS2 is well validated as it incorporates 11 experimental values boosting its robustness. Among the tested models, Random Forest Regression (RFR) provides the most accurate predictions for the expanded dataset. This research positions machine learning and data augmentation as efficient, scalable alternatives to traditional methods, offering a faster and more reliable approach to electronic component lifespan prediction.