The work aims to meet the increasingly personalized customization needs of the corrugated board industry, reduce the impact of complex and variable production conditions on the production speed, help enterprises to arrange production reasonably, and improve the level of production line control. Firstly, the production speed of corrugated board was resampled to unify the sampling interval of order parameters and sensor parameters, and high pass filtering by ButterWorth filter. Quartile statistics were used to screen the stable wet end production speed interval and extract the data of types B and BC. Then, BP neural network and XGBoost were used to predict the wet end production speed based on the extracted data, and Bayesian optimization and grid search were used to optimize the hyperparameter of two models, respectively. Finally, PSO algorithm was used to combine the two models to predict the production speed. The experimental results showed that both models had certain prediction ability, among which XGBoost had better prediction performance and the combined model had the best prediction performance. The method based on BPNN-XGBoost combined model can effectively predict the wet end production speed of corrugated board and guide the actual production.