中密度纤维板施胶系统的神经网络预测控制

Neural Network Predictive Control of the Medium Density Fiberboard Glue System

针对中密度纤维板(MDF)施胶过程的特点,提出一种基于神经网络预测控制原理的施胶控制方法。将MDF施胶系统作为被控对象,利用BP神经网络辨识其模型,并采取多步递推的方式得到被控对象的预测模型。为验证所提方法的控制效果,对MDF施胶系统进行了仿真和实时控制试验,并与PID控制进行比较。结果表明:设计的控制方案,为MDF施胶系统获得良好的控制品质和抗干扰能力,实现施胶量对纤维量动态变化的快速跟踪,比PID控制具有明显的优势。

According to the characteristics of the sizing process of medium density fiberboard （ MDF） , we proposed a gluing con-trol strategy with neural network predictive control principle.By regarding MDF glue system as a control, we used BP neu-ral network to identify model and multi-step recursive to obtain the prediction model .We made some simulations and real-time control experiments for MDF gluing system and validates the effect compared with the PID control.The glue system brings MDF system a good tracking performance and anti-jamming capability, and realizes the fast track of glue consump-tion in dynamic changes of fiber amount with obvious advantages compared with the PID control .