关键词: 电子封装, 异质互连结构, 随机有限元, 数据驱动

Abstract: Heterogeneous interconnects in electronic packaging are crucial for chip miniaturization and functional integration. Their numerical analysis involves multi-source uncertainties, material and geometric nonlinearities, and multi-physics coupling. In recent years, stochastic finite element method (SFEM) simulation has been widely used in electronic packaging, providing an effective tool for uncertainty quantification and reliability assessment. This review concludes the modeling theory and numerical implementation methods under three types of uncertainties: geometric, material, and boundary conditions. We focus on dimensional deviations, warpage, material property fluctuations, and operational perturbations, summarizing current practices combining SFEM, data-driven SFEM, and Kriging surrogate models. By introducing random variables and random fields into the thermo-mechanical-electric coupling model, the research reviewed in this paper achieves high-dimensional uncertainty propagation and statistical reliability assessment, marking a shift from deterministic simulation to probabilistic analysis. We further review data-driven SFEM methods that integrate experimental, standard, and simulation data for regularization and identification of multi-source probabilistic variables; in addition, we review Kriging surrogate models that support high-dimensional nonlinear response prediction and uncertainty quantification. In summary, the reviewed literature provides a scalable theoretical and numerical foundation for the reliability design, lifetime prediction, and digital twin development of complex electronic packaging systems.

Key words: Electronic packaging, heterogeneous interconnects, stochastic finite element method, data-driven