In the preparation of chemical raw materials, there are often some optimization problems that are difficult for people to solve. Due to the complex interaction between various factors and the great difference between the theoretical model and the practice, an effective optimal design model is imminent. In this paper, an innovative mixed statistical model is proposed for optimal design, including five sub-models of ANOVA model, polynomial optimal regression analysis model, design uniformity model, two-step regression model, and optimization model. In this paper, the mixed statistical model is also applied to the study of the combination of catalyst and temperature in the preparation of C4 olefin production, which has high practicability.