The objective of this research is to predict the solder joint’s fillet profile before its formation. Solder joints are crucial for the structural and operational reliability of electronic assemblies, yet their integrity can be compromised by defects such as cold joints, voids, or insufficient solder. Traditional physics-based simulations attempt to model these phenomena but often fall short due to simplifications that fail to capture real-world variability. Conversely, data-driven approaches leverage historical data from Surface Mount Technology (SMT) lines to predict joint quality, though their effectiveness can be hampered by data noise and imbalance. Addressing these limitations, this research introduces a hybrid modeling framework named Sim2Joint, which combines physics knowledge-based simulations with the adaptability of data-driven methods. By introducing Sim2Real in the joint simulation domain, Sim2Joint bridges the gap between simulation and real-world situations by integrating dynamic weights for printing and placing factors with real-world data, enhancing prediction accuracy and reliability. The framework also includes uncertainty quantification to provide more reliable solder joint fillet profile predictions, thereby enabling better decision-making and optimization in SMT processes. Sim2Joint is validated against various baselines, showcasing its capability to adapt to real-time changes and improve the predictive performance of solder joint quality assessments.

中文翻译：

---

Sim2Joint：用于跨 Sim2Real 预测焊点的动态混合模型


本研究的目的是在焊点形成之前预测焊点的圆角轮廓。焊点对于电子组件的结构和操作可靠性至关重要，但其完整性可能会因冷点、空洞或焊料不足等缺陷而受到损害。传统的基于物理场的仿真试图对这些现象进行建模，但由于简化无法捕捉到现实世界的可变性，因此往往达不到要求。相反，数据驱动的方法利用表面贴装技术 （SMT） 生产线的历史数据来预测接头质量，尽管它们的有效性可能会受到数据噪声和不平衡的阻碍。为了解决这些限制，本研究引入了一个名为 Sim2Joint 的混合建模框架，该框架将基于物理知识的模拟与数据驱动方法的适应性相结合。通过在联合仿真领域引入 Sim2Real，Sim2Joint 通过将用于打印和放置因素的动态权重与真实世界数据集成，提高了预测的准确性和可靠性，从而弥合了模拟和真实世界情况之间的差距。该框架还包括不确定性量化，以提供更可靠的焊点圆角轮廓预测，从而在 SMT 流程中实现更好的决策和优化。Sim2Joint 针对各种基线进行了验证，展示了其适应实时变化和提高焊点质量评估预测性能的能力。