The number and distribution of sample points are important factors which affect the accuracy of the Kriging model. The number of sample points should be so sufficient that these points can be spread over the design space uniformly. In this paper, the initial 45 samples were selected by LHS and the performance of 45 samples was evaluated by the commercial code Fluent. The initial Kriging model was then constructed based on this sample data. As already discussed, the objective function L/D was transformed to the EI to avoid missing the global maximum point. The objective functions are expressed as follows: