造纸过程定量水分的小波神经网络PID控制  

PID Control of Basis Weight and Moisture in Paper Making Process Using Wavelet Neural Network

作　　者：李艳[1] 许合金[2] 

机构地区：[1]陕西科技大学电信学院,陕西西安710021 [2]陕西科技大学动力科,陕西西安710021

出　　处：《计算机仿真》2012年第3期249-253,共5页Computer Simulation

基　　金：陕西省教育厅专项科研项目(2010JK426)

摘　　要：抄纸过程定量水分控制系统存在时变、时滞、非线性、不确定等特点,使得定量、水分两个指标波动幅度较大,严重地影响了成纸合格率,断纸次数多。为解决上述问题,设计了定量水分的串级控制方案,采用小波神经网络PID控制方法对定量水分控制,并建立了三层神经网络模型,输出层单元分别表示PID控制器的三个参数,用非线性小波基函数取代隐层神经元的Sigmoid函数,根据定量、水分的实际输出与期望输出之间的偏差可实时在线调整PID参数。在MATLAB上进行仿真。仿真结果表明,控制系统能较好地克服时滞、时变、非线性等干扰对定量水分的影响,实现控制精度高,能很快达到稳定,并可有效地减少断纸次数,提高造纸效率。The features of time-varying,time-delay,nonlinear,uncertainty and others in weight and moisture control in paper making process cause large fluctuations of weight and moisture,so lead to the increasing number of sheet breaks,which affects the utilization and passing rate of paper.In order to decrease the number of sheet breaks,the mathematical model and factors influencing weight and moisture were analyzed,a cascade control scheme was designed,and WNNPID control algorithm was used.Three-layers neural network was created.The input layer units were the desired outputs,actual outputs and the error signals,and the output layer units were the parameters of PID respectively.Nonlinear wavelet function was used to replace sigmoid function of hidden layer neurons,parameters were adjusted on-line according to the errors between expectations and actual values.Simulation results show that this control system can overcome the influences from time-varying,time-delay,nonlinear,uncertainty,interference and etc,and increase the control accuracy and control speed.The practical results show that this control system can reduce the number of sheet breaks greatly and increase the economic efficiency of enterprises.

关 键 词：小波神经网络 参数在线整定 定量 水分 比例积分微分控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]