摘要
This study explores the mechanical and thermal behavior of glass substrates under diverse physical loading conditions. To systematically evaluate the primary factors influencing warpage performance and stress distribution in glass substrate applications, finite element analysis (FEA) is employed and validated through high-precision optical measurements. Furthermore, machine learning (ML) is integrated with FEA to develop a predictive modeling framework for glass substrate applications, particularly in redistribution layers (RDL) and through glass vias (TGVs) in glass interposers. This study investigates the mechanical and thermal behavior of glass substrates under diverse loading conditions relevant to advanced packaging. Finite element analysis (FEA) is employed and validated against high-precision optical measurements to assess warpage and stress distribution. To enhance predictive capabilities, machine learning (ML) is integrated with FEA, particularly focusing on redistribution layers (RDL) and through-glass vias (TGVs). Glass substrates, favored for their high resistivity and low dielectric loss, are increasingly utilized in high-speed and RF applications. However, their brittle nature introduces challenges in structural reliability. This work combines calibrated FEA datasets and support vector algorithms (SVA) to develop an accurate predictive model (R2 = 0.9721) and further compares its performance against Decision Tree, Random Forest, and Artificial Neural Network models. The study also analyzes thermo-mechanical failure mechanisms in TGVs under multi-physics loading, establishing a data-driven simulation workflow for robust reliability evaluation in glass-based packaging architectures. Copyright © 2025 by ASME.