Machine Learning based Prediction of Wire Bonding Profile in 3D stacked integrated microelectronic packaging

Wire bonding (WB) have been widely used in 3D stacked integrated microelectronic packaging because of its high reliability and low-cost. In 3D stacked packaging, the required size for microelectronic devices was much smaller than before, which led the shape of wire profile becoming more critical as it directly affects the total thickness of the chip package. However, there is no facile means to predict the wire profile. In this study, machine learning (ML) was proposed to predict the wire profile of wire bonding under various conditions. After collecting abundant relevant information using finite element simulation experiments, the support vector regression (SVR) was selected as the predictor while support vector machine (SVM) was used as the classifier to the train the ML model, thereby, the wire bonding profile that can meet the requirements of microelectronic industry can be obtained. This study can provide useful insights for developing wire bonding processing techniques.