A Hybrid Model for Solder Joint Height Prediction Based on Physical Knowledge and Machine Learning

This study presents a hybrid model for solder joint profile calculation before formation, merging physics-based insights with machine learning techniques. As the electronics industry shifts towards miniaturization and functionality, solder joint quality emerges as a vital factor for ensuring the electrical and mechanical integrity of electronic assemblies. The ability to predict solder joint height prior to the reflow process is key for real-time quality control. This study develops a numerical simulation model to predict the cross-sectional height of solder joints based on how molten solder wets the surfaces of pads and chip terminations during reflow. To align theoretical models with actual outcomes, machine learning is employed to refine prediction results by establishing correlations between chip placement and the cross-sectional area of solder joints. This method not only offers accurate predictions of solder joint height quickly and simply but also provides model interpretability, thus improving decision-making in the manufacturing process.