Abstract: In order to improve the prediction accuracy of a kriging model, an improved adaptive infilling criterion and a parallel infilling strategy are proposed based on the maximum expected improvement criterion. Based on the prediction response and prediction variance provided by a kriging model, multiple new points are obtained to update a kriging model by solving multiple infilling criteria in parallel through the particle swarm optimization algorithm, greatly improving the modeling efficiency. The proposed strategy is applied to low-dimensional and high-dimensional classical nonlinear functions and compared with the single-point infilling method. The results show that the proposed strategy can reduce sampling times by more than 50% at the same global accuracy, and the total number of samples required for a kriging model is less. Finally, the proposed strategy is applied into the sheet metal forming, and the fender in the NUMISHEET2002 is taken as the research object. The kriging model between the equivalent drawbead restraining force and the thinning rate is established to obtain the optimal solution of the drawbead resistance by parallel infilling strategy. So the cracking defect is completely eliminated in the fender forming process and the forming quality of the blank is improved. The research shows that the improved adaptive infilling criterion and the proposed parallel infilling strategy can effectively improve the modeling efficiency and prediction accuracy of the kriging model.

Key words: drawbeads, kriging model, parallel infilling criterion, sheet metal forming