神经网络在液位控制系统研究中的应用

液位控制系统是典型的一阶系统 ,实现该系统控制通常用浮子检测控制方法 鉴于其存在的不足 ,文中提出了神经网络控制方法 ,即通过神经网络的自学习功能和泛化能力 ,达到其能取代浮子检测而控制流量阀实现输入流量的控制 ,通过压强与流阻的乘积关系而自行调节控制出口流量 同时 ,本文研究中间层层数、权重及学习速率对神经网络的学习训练时间产生的影响 ,并在此基础上 ,选择合适的中间层层数、权重、学习速率 ,使神经网络对该液位控制系统进行有效的学习

The Application of Neural Networks in the System of Liquid Height

Liquid height control system belongs to a typical one step system. The usual approach to control it is with buoyage measuring. In terms of its drawbacks, the paper proposes the neural networks control. That is to say, neural networks may adjust flux valve to control input flux instead of buoyage measuring because of its self study and generalization. However, output flux can adjust and control automatically through the product of pressure and resistance. At the same time, the effect of hidden neurons, momentum and learning rate on the learning time is studied. Then choosing the appropriate hidden neurons, momentum and learning rate is of vital importance for neural networks to study the liguid height control system efficiently.