基于改进的神经网络最近邻聚类算法在结晶器液位系统中的应用  

Application of the improved neural network nearest neighbor clustering algorithm in mold level system

作　　者：缸明义[1] 陈立辛 宁平华[1] 夏兴国[1] GANG Ming-yi;CHEN Li-xin;NING Ping-hua;XIA Xing-guo(Department of Electrical Engineering,Maanshan Technical College,Anhui Maanshan 243031,China;College of Mechanical Engineering,Anhui Science and Technology University,Anhui Fengyang 233100,China)

机构地区：[1]马鞍山职业技术学院电气工程系,安徽马鞍山243031 [2]安徽科技学院机械工程学院,安徽凤阳233100

出　　处：《齐齐哈尔大学学报（自然科学版）》2021年第2期23-28,共6页Journal of Qiqihar University(Natural Science Edition)

基　　金：安徽高校自然科学研究重点项目(KJ2019A1245,KJ2019A1244);安徽省高校优秀青年人才支持计划项目(gxyqZD2018105,gxyq2019202)。

摘　　要：针对连铸机的结晶液位采用拉速控制导致控制过程不稳定而影响铸坯质量的问题,提出了基于神经网络的模型辨识及智能PID控制方法,它主要基于径向基函数(即RBF)神经网络,通过改进的最近邻聚类学习算法在线辨识相关的结晶器系统模型。基于径向基函数辨识网络,将辨识所得雅克比阵应用到智能PID控制器的权值调整之中。结果表明,该算法可对结晶器液位控制方面的主要问题进行良好的解决,其适用性已经得到了仿真结果的充分验证。The model identification and intelligent PID control method based on neural network are proposed.It is mainly based on radial basis function(RBF)neural network to identify relevant mold system models online through improved nearest neighbor clustering learning algorithm.Based on the radial basis function identification network,the identified matrix is applied to the weight adjustment of intelligent PID controller.The results show that the algorithm can solve the main problems of mold level control well,and its applicability has been fully verified by simulation results.

关 键 词：最近邻聚类 结晶器 神经网络PID 液位系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]