Abstract: IMEGO algorithm proposed to conquer premature convergence might occur in EGO algorithm when the objective function varied in a large range in this study. The optimal solution of the kriging model for each iteration was preferred and stopping criteria on expected improvement was added in the IMEGO algorithm. The IEGO was compared with traditional EGO through a onevariable test function and three twovariables test functions, it was then shown that the improved EGO could effectively search the global optimum. The IMEGO was used to optimize injection molding process parameters for covers of liquid crystal display parts and the results showed that it had a good application in engineering.

Key words: injection molding, efficient global optimization, expected improvement