基于神经网络逆模型的残碱控制系统  

Residual Soda Control System Based on Neural Network Inverse

作　　者：贺庆之[1] 贺静 

机构地区：[1]陕西科技大学,陕西咸阳712081 [2]飞利浦电子元件(上海)有限公司,上海200131

出　　处：《中国造纸学报》2002年第2期97-100,共4页Transactions of China Pulp and Paper

摘　　要：讨论了一个比较复杂的纸浆洗涤过程的数学模型 ,经过多次搜寻 ,可用五层串并联叠层BP神经网络来模拟这种多输入单输出高价模型 ,给出了该生产过程的神经网络逆模型结构及主要内部特征 ,以及用该神经网络逆模型进行仿真的情况。并运用此神经网络逆模型 ,设计出一套基于神经网络逆模型的残碱开环控制系统 ,再附加一个残碱串级反馈控制作为主控系统的辅助监督控制 ,或主控故障时的备用控制。A mathematical model of pulp wash process was discussed, this multi-input and single-output model could be simulated by five layers Cascade-Forward backpropagation neutral network. The neural network inverse model structure, its internal characteristic and to conduct simulation by using this model were presented. A residual soda open loop control system was designed based on neural network inverse mathematical model and addition superintendence control system of residual soda series class feedback control or reservation control system. In this system, on line collection dynamic data of controlled target inputs and output, use them off line to debug structure of neural network inverse mathematical model, and on line recursive training neural network inverse model, update weighting values W and threshold values b, automatic adaptation changes of target characteristic.

关 键 词：残碱控制系统 神经网络逆模型 辅助监督控制 纸浆 冼涤 质量指标 

分 类 号：TS736.2[轻工技术与工程—制浆造纸工程] TS74