Effective parametric optimization for packaging design using Bayesian optimization

As different structure, materials and components are widely applied in electronic packaging, complicated responses of these heterogenous but integrated parts subjecting to thermo-mechanical loading during manufacturing and usage are becoming a crucial issue to reliability of packaging. Therefore, how to find optimal engineering design of packaging receives growing attention. For the past, Design of Experiment (DoE) combined with statistical analysis is generally adopted. However, the increasing complexity of electronic packaging dramatically enlarges the amount of design parameters, making these traditional methods hard to maintain its efficiency. In current study, we implement Bayesian optimization with Gaussian process (BO-GP) as a framework to search the optimal combination of design parameters to reduce the warpage of a testing packaging vehicle. Upper confidence bound (UCB) method is adopted to define how this framework search in the design space. By employing simulated annealing algorithm, the BO-GP framework in this study can balance the exploration and exploitation within the space and present good convergence towards optimal design parameters.