Due to the excessive dependence on the real time measurement of dominant variables during prediction, it is difficult to apply the nonlinear autoregressive with exogenous input(NARX) model in real industrial processes.In order to solve the difficulty, a new NARX model structure is built by combining two network models.When the industrial environment cannot provide the history information of dominant variables in real time, the real-time prediction of dominant variables can be guaranteed by using the multi-step prediction method.By reducing the autocorrelation of prediction sequence, the cumulative error caused by multi-step estimation is suppressed. At the cost of reducing the accuracy of single-step prediction properly, the model can obtain better prediction effect in the much rough industrial environment.Simulation results verify the effectiveness of the proposed model.