In order to improve the control accuracy, this article proposed a new mathematical model of flow field. By using NARMAX model as system model structure, combining characteristics of the wind tunnel operation, divided the whole process into three independent steps, model of each step was established respectively, and the wind tunnel model was acquired by using function estimation. The order of the model and the sampling interval was identified by the pseudo nearest neighbor method and mutual information method respectively. BP neural network was used for mapping the network input to output as a nonlinear function, the total pressure in settling chamber and the static pressure in plenum chamber can be obtained by utilizing this model, and the Mach number was computed by using the Mach number calculation formula. The result shows that the model has high accuracy of system identification, and is able to embody characteristics of the operation of transonic wind tunnels.