Surrogate models are often used to replace expensive simulation models in complicated engineering problems. The common practice is to construct multiple metamodels based on a common training data set, evaluate the accuracy, and then use only a single model perceived as the best while discarding the rest. This practice has some shortcomings as it does not take full advantage of the resources devoted to constructing different metamodels and increases the risk of adopting an inappropriate model. However, ensemble technique is an effective way to make up for the shortfalls of traditional strategy. In order to improve the efficiency, accuracy and robustness of the surrogate model, an optimal iterative weight factors method for constructing ensemble of surrogate model was proposed. At first, the leaveoneout cross validation strategy and PRESS criterion were presented to calculate initial weight factors. Then, an iterative process for the weight factors was conducted and at the same time the weight factors were updated until an ideal prediction accuracy of the ultimate ensemble of surrogate model was reached. To evaluate the effectiveness of the proposed method, three metamodels and five ensembles of surrogate model for three benchmark problems and an engineering problem were constructed to compare their performances of efficiency, accuracy and robustness. Results show that the proposed method can not only get a higher accuracy and robustness surrogate, but also shorten the time of constructing surrogate model evidently.