In the actual operation process，the dynamic characteristics of the cutter suction dredger （CSD） are very complicated，and many factors affect the production. It will consume much time if all these control factors participate in the output prediction. To train the network and predict the production in real-time，the PCA is utilized to these control factors. Then，the number of control factors is reduced according to the results. In the system simulation modeling，the whole control factors and reduced control factors are used separately as the input variables of the RBF neural network，and the production is used as the output variable to establish the prediction model of CSD’s production. The results show that the prediction model of production can be simplified and the computation time can be reduced while the input variables are reduced，and the prediction accuracy of the model can be maintained，providing real-time production reference for operators on the construction site.