BP神经网络在秸秆挤压机中的应用

Application of BP Neural Network in Straw Extruder

以秸秆挤压机的秸秆含水率、螺杆转速、模孔间隙、套筒温度、螺杆末端至模板内表面的距离(δ段长度)等5种过程参数为输入,以挤压膨化后玉米秸秆纤维含量变化(本文研究NDF与ADF)为输出,依据五因素五水平(1/2实施)二次正交旋转组合试验设计及试验数据建立了系统的BP神经网络模型。该网络训练后得到挤压系统的自变量与因变量之间的映射关系,且具有较好的仿真精度,可实现对过程参数的控制和目标输出的预测,用于指导生产实践。

According to five factors and five levels（1/2 implementation） two orthogonal rotation mix experimentation design and experimentation data, a BP neural network model is builded by serving the straw extruder＇s five process parameters as the input, i.e. materials wet rates, variable rotational speed, template＇s hole gap, sleeve temperatures and the length of the Delta, and by serving the change of fiber＇s contents （here refers to NDF and ADF） as output. After network training, we could gain the mapped relationship between independent variable and dependent variable in the extrusion system. The network model has better imitated precision. It can carry out the control of process parameters and the prediction of goal output and be used for supervising manufacture and practice.