Prediction and Optimization of Residual Stress After Reflow Soldering of BGA Solder Joints Based on Multi-fidelity Surrogate Model

To solve the problem of difficult physical measurement and high simulation cost of residual stress after reflow soldering of BGA solder joints, a prediction and optimization method of the maximum residual stress of BGA solder joints based on the Co-Kriging multi-fidelity surrogate model (MFS) is proposed. The maximum residual stress after reflow soldering is obtained by simulating the melting forming and cooling solidification stages of the solder joints. The MFS between the size parameters such as the height of the solder joint and the radius of the pad and the maximum residual stress after reflow soldering is established. The meshes of different sizes are divided into two stages, and the nested and non-nested sampling methods are adopted to form four MFSs. By comparing the prediction effect of the MFS with the single-fidelity surrogate model, namely the radial basis function, the particle swarm optimization (PSO) algorithm is used to optimize the solder joint size parameters based on the prediction results of the MFS. The results show that the prediction effect of the MFS is better than that of the single-fidelity surrogate model; The best construction scheme of the MFS is: in the simulation stage of solder joint melting forming, mesh generation with different precision is carried out to obtain high/low precision sample points, and the sample points are not nested. The results of PSO prove the validity of the MFS and obtain the optimal solder joint parameters.