Dynamic Adaptive Variable-Fidelity Modeling Method for BGA Solder Joint Design

Currently, building variable-fidelity surrogate models and using dynamic adaptive sampling is becoming a hot research topic in the field of surrogate-based design (SBD). However, due to its unique sample composition, the variable-fidelity model not only needs to consider the acquisition method of sampling points but also the number of samples with different fidelity levels. This poses higher requirements for dynamic adaptive sampling strategies. In this paper, we propose a co-kriging sequential modeling method based on SOM and LHD, which uses the space volume determined by SOM to obtain the sampling number of different fidelity samples. The feasibility of the proposed method was verified using the reliability problem of BGA soldering. Comparing the results obtained by the one-time modeling method for this problem, we discussed the advantages and disadvantages of this approach.