The BP neural network model is established based on the design theory of traditional twin-screw extrusion machine and the property of wood-plastic composite. The relationship of screw’s diameter and rotating speed of twin-screw extruder special for wood-plastic composite material are predicted. Firstly, taking the viscosity of wood plastic composite material, the pressure of extruder’s head, the temperature of screw’s metering section and the target yield as input variables, the screw’s diameter and rotating speed as output variables and the design theory of traditional double screw extrusion machine as the momentum equation, establishes the BP neural network model. Then, through the sample inputting, numerical fitting training of the model is conducted until the error requirement is met. Finally the model is applied to forecast the diameter and rotating speed of twin-crew, and the optimal result is output. The result shows that combining with traditional design theory, the intelligent network model established through material properties and yield integrated sample input can better simulate the complicate conditions of screw motion in practical production.