Application of artificial neural network in thermal and solder joint reliability analysis for stacked dies LBGA

Thermal analysis and solder joint reliability (SJR) analysis in electronic is very crucial during the design stage. Finite element method (FEM) becomes popular in predicting the thermal and SJR performance of electronic packaging due to expensive and laborious experiment setup. However, FEM involves complex theory of physic and mathematic modeling with tedious material properties, meshing and boundary condition setup which required experts and long computational time. Artificial neural network (ANN) is an alternative tool to predict thermal and SJR performance of electronic packages if the historical data for training is available. The trained ANN is user friendly, fast and accurate tool to predict the thermal and SJR performance of electronic packages during the design stage. This paper will discuss about FEM procedure which is used to produce training data for ANN. The packages used in the study are LBGA stacked dies which gaining popularity in recent years due to the enabling of integration of multiple system and subsystem into one package. The results of thermal and SJR analysis which were predicted ANN agreed well with the FEM result and data from publications.