Stress–strain analysis and prediction of WLCSP solder joints under bending–torsion coupled loading

With the trend of miniaturization in electronic components, Wafer-Level Chip Scale Package(WLCSP) chips, known for their compact size, are widely used in various electronic devices. During the operation of these devices, they are inevitably subjected to bending and torsional loads, such as frequent touches on smartphone screens, the opening and closing actions of laptop screens, and keyboard typing. Research on the stress and strain of WLCSP solder joints is crucial for the reliability design of these joints, thereby enhancing the overall reliability of electronic devices. Therefore, this study establishes a finite element model of WLCSP solder joints and conducts a finite element analysis of their stress and strain under bending–torsion coupling loads.A strain testing platform for WLCSP solder joints under bending–torsion coupled loading was developed, and strain measurement tests were performed to verify the accuracy of the simulation results. Three structural parameters of the WLCSP solder joint — solder joint diameter, pad diameter, and solder joint height — were selected as influencing factors. Using the orthogonal design method, 16 different combinations of structural parameter levels for the WLCSP solder joints were designed to obtain the maximum bending–torsion coupled stress, followed by range and variance analysis of the maximum stress. A particle swarm optimization (PSO)–Back Propagation(BP) neural network prediction model for the bending–torsion coupled stress of the WLCSP solder joint was developed to predict the stress. The results indicate that, among the three structural parameters — solder joint diameter, pad diameter, and solder joint height — the ranking of their influence on the WLCSP solder joint bending–torsion coupled stress is as follows: solder joint height $>$ solder joint diameter $>$ pad diameter. When the confidence level is 95%, solder joint height and solder joint diameter have a significant impact on the bending–torsion coupled stress of the WLCSP solder joint. The established PSO-BP neural network model for predicting the bending–torsion coupled stress of the WLCSP solder joint shows a maximum prediction error of 5.58% and an average error of 3.01%, demonstrating a good prediction accuracy for the stress.